BEGIN:VCALENDAR
VERSION:2.0
METHOD:PUBLISH
X-WR-CALNAME: Calendar | UW-Madison Computer Sciences
PRODID:-//Drupal iCal API//EN
BEGIN:VEVENT
UID:calendar.407.field_date.0.0
SUMMARY:Faculty Lunch
DTSTAMP:20130618T164412Z
DTSTART:20110126T180000Z
DTEND:20110126T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/faculty-lunch
DESCRIPTION:CS Faculty Lunch
END:VEVENT
BEGIN:VEVENT
UID:calendar.6916.field_date.0.1
SUMMARY:Scott Davidoff talk
DTSTAMP:20130618T164412Z
DTSTART:20110901T203000Z
DTEND:20110901T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/scott-davidoff-talk
LOCATION:1240 CS
END:VEVENT
BEGIN:VEVENT
UID:calendar.7902.field_date.0.2
SUMMARY:Yupu Zhang Preliminary Exam
DTSTAMP:20130618T164412Z
DTSTART:20111109T180000Z
DTEND:20111109T193000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/yupu-zhang-preliminary-exam
LOCATION:3310 CS
DESCRIPTION:Providing End-to-end Data Integrity in File Systems and Home -user  \n ApplicationsYupu Zhang Computer Sciences\, UW-Madison Committee: Andrea  \n Arpaci-Dusseau & Remzi Arpaci-Dusseau(co-advisors)\; Shan Lu
END:VEVENT
BEGIN:VEVENT
UID:calendar.7903.field_date.0.3
SUMMARY:Architecture Seminar: Asynchronous Computer Arithmetic
DTSTAMP:20130618T164412Z
DTSTART:20111110T170000Z
DTEND:20111110T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/architecture-seminar-asynchronous-computer-arithmetic
LOCATION:1240 CS
DESCRIPTION: Asynchronous Computer ArithmeticRajit Manohar Thursday\, November 10\,  \n 2011 11:00 AM\, 1240 CS(Cookies: 10:30)Abstract: The time and energy  \n complexity of arithmetic operations are highly dependent on their input  \n operands. Synchronous implementations tend to treat all input scenarios  \n uniformly due to a uniform clock cycle period constraint. In this talk I will  \n show how asynchronous circuits can exploit operand-dependent behavior to  \n provide implementations of common arithmetic operations that are\, on average\,  \n more efficient than conventional solutions.
END:VEVENT
BEGIN:VEVENT
UID:calendar.7917.field_date.0.4
SUMMARY:SEPP: SATe-Enabled Phylogenetic Placement
DTSTAMP:20130618T164412Z
DTSTART:20111111T180000Z
DTEND:20111111T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/sepp-sate-enabled-phylogenetic-placement
LOCATION:140 Bardeen
DESCRIPTION: Tandy Warnow David Bruton\, Jr. Centennial Professor\, Department of  \n Computer Sciences\, University of Texas at Austin  Phylogenetic placement  \n arises in the analysis of metagenomic data\, in which the objective is to  \n insert short molecular sequences (called "query sequences") into an existing  \n phylogenetic tree and alignment on full-length sequences for the same gene.  \n Phylogenetic placement has the potential to provide information beyond pure  \n species identification (i.e\, the association of metagenomic reads to existing  \n species)\, because it can also give information about the evolutionary  \n relationships between these query sequences and to known species. We present  \n SEPP\, a general "boosting" technique to improve the accuracy and/or speed of  \n phylogenetic placement techniques. The key algorithmic aspect of this booster  \n is a dataset decomposition technique in SATe (Liu et al.\, Science 2009)\, a  \n method that utilizes an iterative divide-and-conquer technique to co-estimate  \n alignments and trees on large molecular sequence datasets. We show that SEPP  \n improves current phylogenetic placement methods\, placing metagenomic  \n sequences more accurately when the set of input sequences has a large  \n evolutionary diameter and produces placements of comparable accuracy in a  \n fraction of the time for easier cases. Joint work with Siavash Mirarab and  \n Nam Nguyen\, PhD students at UT-Austin.
END:VEVENT
BEGIN:VEVENT
UID:calendar.7918.field_date.0.5
SUMMARY:Practical Verified Computation with Streaming Interactive Proofs
DTSTAMP:20130618T164412Z
DTSTART:20111111T190000Z
DTEND:20111111T200000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/practical-verified-computation-streaming-interactive-proofs
LOCATION:4310 CS
DESCRIPTION: Michael Mitzenmacher Department of Computer Science\, Harvard  \n University Friday\, November 11\, 2011 1:00 PM\, 4310 CS(Cookies: 1:00)A  \n potential problem in outsourcing work to commercial cloud computing services  \n is trust. If we store a large data set with a service provider\, and ask them  \n to perform a computation on that data set -- for example\, to compute the  \n eigenvalues of a large graph\, or to compute a linear program on a large  \n matrix derived from a database -- how can we know the computation was  \n performed correctly? Obviously we don't want to compute the result ourselves\,  \n and we might not even be able to store all the data locally. This leads to  \n new problems in the streaming paradigm: we consider a streaming algorithm  \n (modelling a user with limited memory and computational resources) that can  \n be assisted by a powerful helper (the service provider). The goal of the  \n service provider is to not only provide the user with answer\, but to convince  \n the user the answer is correct.In this talk\, I will give a unified overview  \n of a recent line of work exploring the application of proof systems to  \n problems that are streaming in nature. In all of these protocols\, an honest  \n service provider can always convince the data owner that the answer is  \n correct\, while a dishonest prover will be caught with high probability. The  \n protocols I will discuss utilize and extend powerful ideas from communication  \n complexity and the theory of interactive proofs\, and I will argue that many  \n are highly practical\, achieving millions of updates per second and requiring  \n little space and communication.Joint work with Graham Cormode and Justin  \n Thaler.
END:VEVENT
BEGIN:VEVENT
UID:calendar.7919.field_date.0.6
SUMMARY:Optimal Newton-type methods for nonconvex smooth optimization
DTSTAMP:20130618T164412Z
DTSTART:20111114T220000Z
DTEND:20111114T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/optimal-newton-type-methods-nonconvex-smooth-optimization
LOCATION:Room 3280B 3rd floor\,...
DESCRIPTION:Coralia Cartis Lecturer\, School Of Mathematics\, University of Edinburgh\,  \n United Kingdom Monday\, November 14\, 2011 4:00 PM\, Wisconsin Institute for  \n Discovery (WID)\, Room 3280B 3rd floor\, teaching lab)\, Anyone without WID  \n access can use the special events elevator on the WID 1st floor (near Aldo  \n coffee shop) to access the WID 3rd floor teaching lab.We show that the  \n steepest-descent and Newton's methods for unconstrained nonconvex  \n optimization under standard assumptions may both require a number of  \n iterations and function evaluations arbitrarily close to the  \n steepest-descent's global worst-case complexity bound. This implies that the  \n latter upper bound is essentially tight for steepest descent and that  \n Newton's method may be as slow as the steepest-descent method in the worst  \n case. Then the cubic regularization of Newton's method (Griewank (1981)\,  \n Nesterov & Polyak (2006)) is considered and extended to large-scale problems\,  \n while preserving the same order of its improved worst-case complexity (by  \n comparison to that of steepest-descent)\; this improved worst-case bound is  \n also proved to be essentially tight. We further show that the cubic  \n regularization approach is\, in fact\, optimal from a worst-case complexity  \n point of view amongst a class of second-order methods. Time permitting\, the  \n problem-evaluation complexity of constrained optimization will also be  \n discussed\, and shown to have\, surprisingly\, the same order as a function of  \n the accuracy as in the unconstrained case. This is joint work with Nick Gould  \n (Rutherford Appleton Laboratory\, UK) and Philippe Toint (University of Namur\,  \n Belgium).
END:VEVENT
BEGIN:VEVENT
UID:calendar.7946.field_date.0.7
SUMMARY:AI Seminar: Rebecca Willett
DTSTAMP:20130618T164412Z
DTSTART:20111117T200000Z
DTEND:20111117T210000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/ai-seminar-rebecca-willett
LOCATION:4610 Eng
DESCRIPTION:Scalable Tracking of Dynamic NetworksRebecca Willett\, Assistant Professor\,  \n Electrical and Computer Engineering\, Duke UniversityAbstract: Online  \n optimization methods are useful in a variety of applications with sequential  \n observations of a dynamic environment. Often such methods are designed to  \n minimize an accumulated loss metric\, and the analysis techniques are  \n appealing because of their applicability in settings where observations  \n cannot be considered independent or identically distributed\, and accurate  \n knowledge of the environmental dynamics cannot be assumed. However\, such  \n analyses may mask the role of regularization and adaptivity to environmental  \n changes. This work explores regularized online optimization methods and  \n presents several novel performance bounds. Tracking regret bounds relate the  \n accumulated loss of such an algorithm with that of the best possible dynamic  \n estimate that could be chosen in a batch setting\, and risk bounds quantify  \n the roles of both the regularizer and the variability of the (unknown)  \n dynamic environment. The efficacy of the method is demonstrated in an online  \n Ising model selection context applied to U.S. Senate voting data.
END:VEVENT
BEGIN:VEVENT
UID:calendar.7948.field_date.0.8
SUMMARY:AI Seminar: Dewey\, Li\, & LeGault
DTSTAMP:20130618T164412Z
DTSTART:20111117T220000Z
DTEND:20111117T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/ai-seminar-dewey-li-legault
LOCATION:3310 CS
DESCRIPTION:Meet the AI GroupProf. Colin Dewey\, Bo Li and Laura LeGaultAbstract: Come  \n meet members of the UW Artificial Intelligence community! This is a chance  \n for new and returning students to get a feel for AI research being done here  \n at UW. It is also an opportunity to introduce students to professors and  \n researchers they may want to work with in the future. This week with feature  \n talks from Prof. Colin Dewey and two of his students\, Bo Li and Laura Hobbes  \n Legault. Questions\, contact bgibson@cs.wisc.edu Light refreshments will be  \n served.
END:VEVENT
BEGIN:VEVENT
UID:calendar.7950.field_date.0.9
SUMMARY:SACM Outreach: Research Spotlight
DTSTAMP:20130618T164412Z
DTSTART:20111118T213000Z
DTEND:20111118T223000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/sacm-outreach-research-spotlight
LOCATION:1240 CS
DESCRIPTION:SACM Research SpotlightAbstract: Graduate students from a variety of areas in  \n the department will speak on their current research. The featured speakers  \n are: Dan Szafir\, Human Computer Interaction\; Evan Driscoll\, Programming  \n Languages\; Lance Hartung\, Networking\; Spyros Blanas\, Databases\; Danielle  \n Albers\, Graphics/Vision\; Adrian Mayorga\, Graphics/Vision. This is a great  \n opportunity for graduate students at all levels to learn about some of the  \n exciting research going on in the department. All are encouraged to attend  \n TGIF at 4:30pm in CS 2310 following this event.
END:VEVENT
BEGIN:VEVENT
UID:calendar.7949.field_date.0.10
SUMMARY:Math Colloquium: Recht
DTSTAMP:20130618T164412Z
DTSTART:20111118T220000Z
DTEND:20111118T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/math-colloquium-recht
LOCATION:B239 Van Vleck
DESCRIPTION:The Convex Geometry Of Inverse ProblemsBenjamin Recht\, Assistant Professor\,  \n Department of Computer Sciences\, University of Wisconsin--MadisonAbstract:  \n Deducing the state or structure of a system from partial\, noisy measurements  \n is a fundamental task throughout the sciences and engineering. The resulting  \n inverse problems are often ill-posed because there are fewer measurements  \n available than the ambient dimension of the model to be estimated. In  \n practice\, however\, many interesting signals or models contain few degrees of  \n freedom relative to their ambient dimension: a small number of genes may  \n constitute the signature of a disease\, very few parameters may specify the  \n correlation structure of a time series\, or a sparse collection of geometric  \n constraints may determine a molecular configuration. Discovering\, leveraging\,  \n or recognizing such low-dimensional structure plays an important role in  \n making inverse problems well-posed. In this talk\, I will propose a unified  \n approach to transform notions of simplicity and latent low dimensionality  \n into convex penalty functions. This approach builds on the success of  \n generalizing compressed sensing to matrix completion\, and greatly extends the  \n catalog of objects and structures that can be recovered from partial  \n information. I will focus on a suite of data analysis algorithms designed to  \n decompose general signals into sums of atoms from a simple---but not  \n necessarily discrete---set. These algorithms are derived in a convex  \n optimization framework that encompasses previous methods based on l1-norm  \n minimization and nuclear norm minimization for recovering sparse vectors and  \n low-rank matrices. I will provide sharp estimates of the number of generic  \n measurements required for exact and robust recovery of a variety of  \n structured models. I will then detail several example applications and  \n describe how to scale the corresponding inference algorithms to massive data  \n sets.
END:VEVENT
BEGIN:VEVENT
UID:calendar.7920.field_date.0.11
SUMMARY:Can the Theory of Algorithms Ratify the “Invisible Hand of the Market”?
DTSTAMP:20130618T164412Z
DTSTART:20111121T220000Z
DTEND:20111121T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/can-theory-algorithms-ratify-%E2%80%9Cinvisible-hand-market%E2%80%9D
LOCATION:1240 CS
DESCRIPTION: Prof. Vijay Vazirani College of Computing\, Georgia Institute of  \n Technology Monday\, November 21\, 2011 4:00 PM\, 1240 CS(Cookies: 3:30)“It  \n is not from the benevolence of the butcher\, the brewer\, or the baker\, that we  \n expect our dinner\, but from their regard for their own interest.” Each  \n participant in a competitive economy is “led by an invisible hand to  \n promote an end which was no part of his intention.” -- Adam Smith\,  \n 1776.With his treatise\, The Wealth of Nations\, 1776\, Adam Smith initiated the  \n field of economics\, and his famous quote provided this field with its central  \n guiding principle. The pioneering work of Walras (1874) gave a mathematical  \n formulation for this statement\, using his notion of market equilibrium\, and  \n opened up the possibility of a formal ratification. Mathematical ratification  \n came with the celebrated Arrow-Debreu Theorem (1954)\, which established  \n existence of equilibrium in a very general model of the economy\; however\, an  \n efficient mechanism for finding an equilibrium has remained elusive.The  \n latter question can clearly benefit from the powerful tools of modern  \n complexity theory and algorithms. We will provide an in-depth overview of the  \n fascinating theory that has emerged around this question over the last  \n decade.A compelling new issue is extending this deep understanding of markets  \n to the digital economy -- because of some fundamental reasons\, the  \n methodology outlined above does not carry over to the digital realm. We will  \n outline recent progress on this issue as  \n well.----------------------------------------------------------------------------------Speaker's  \n bio:Vijay Vazirani got his Bachelor's degree in Computer Science from MIT in  \n 1979 and his Ph.D. from the University of California at Berkeley in 1983. His  \n research has spanned a broad range of themes within the design of efficient  \n algorithms - combinatorial optimization\, approximation algorithms\, randomized  \n algorithms\, parallel algorithms\, and most recently algorithmic issues in game  \n theory and mathematical economics. He has also worked in complexity theory\,  \n cryptography and information theory. In 2001 he published what is widely  \n regarded as the definitive book on Approximation Algorithms. This book has  \n been translated into Japanese\, Polish\, French and Chinese. In 2007\, he  \n co-edited a comprehensive volume on Algorithmic Game Theory. During 2011-12  \n he is visiting Stanford University and Caltech under a Guggenheim Fellowship.
END:VEVENT
BEGIN:VEVENT
UID:calendar.7960.field_date.0.12
SUMMARY:CANCELLED Systems Biology: Investigating Modern-Day Disease by Using  \n Metabolic Models of Bacteria from the Past
DTSTAMP:20130618T164412Z
DTSTART:20111122T220000Z
DTEND:20111122T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/cancelled-systems-biology-investigating-modern-day-disease-using-metabolic-models-bacteria-pas
DESCRIPTION:***This talk has been cancelled due to illness.*** Computational Biology  \n SeminarDavid J. Baumler\, PhDGenome Center of Wisconsin UW-Madison
END:VEVENT
BEGIN:VEVENT
UID:calendar.7963.field_date.0.13
SUMMARY:Dynamite: Dynamic Instantiation of Virtual Caching Appliances
DTSTAMP:20130618T164412Z
DTSTART:20111128T170000Z
DTEND:20111128T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/dynamite-dynamic-instantiation-virtual-caching-appliances
LOCATION:2310 CS
DESCRIPTION:One of the key challenges in the data center today is the efficient use of  \n data-center resources while providing services with service-level objectives  \n (SLOs). The primary reason for the challenge is the dynamism in workload  \n requirements over time. We propose dynamic instantiation of virtual caching  \n appliances (caches as virtual machines) to handle the dynamism in workloads  \n and thereby support storage SLOs efficiently. We have developed an SLO-based  \n automation framework called Dynamite for cache instantiation\, that includes  \n low-overhead techniques to: determine the workloads that would benefit from  \n caching\, determine the appropriate cache size for these workloads\,  \n instantiate the cache and non-disruptively migrate the application\, and  \n finally warm the cache to quickly return to acceptable service levels. We  \n have evaluated the effectiveness of the individual techniques using a variety  \n of I/O traces and find that they are highly accurate despite approximations  \n that significantly reduce monitoring overheads. And finally\, the approach  \n actually works! Using the complete pipeline on a case study involving  \n interfering workloads shows that service levels can be met while utilizing  \n resources efficiently.Bio: Lakshmi N. Bairavasundaram is a member of  \n technical staff in the Advanced Technology Group at NetApp. His research  \n interests include storage systems\, file systems\, storage and data management\,  \n and fault tolerance. Lakshmi currently focuses on storage and data management  \n abstractions and techniques. He joined NetApp after completing his Ph.D. in  \n Computer Sciences (2008) at the University of Wisconsin-Madison under the  \n supervision of Prof. Andrea Arpaci-Dusseau and Prof. Remzi Arpaci-Dusseau.
END:VEVENT
BEGIN:VEVENT
UID:calendar.7973.field_date.0.14
SUMMARY:General and Nested Wiberg Minimization
DTSTAMP:20130618T164412Z
DTSTART:20111128T170000Z
DTEND:20111128T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/general-and-nested-wiberg-minimization
LOCATION:1170 WID
DESCRIPTION:Abstract:  Wiberg matrix factorization breaks a matrix Y into low-rank  \n factors U and V by solving for V in closed form given U\, linearizing V(U)  \n about U\, and iteratively minimizing ||Y - UV(U)||_2 with respect to U only.  \n This approach factors the matrix while effectively removing V from the  \n minimization. Recently Eriksson and van den Hengel extended this approach to  \n L1\, minimizing ||Y - UV(U)||_1. We generalize their approach beyond  \n factorization to minimize an arbitrary function that is nonlinear in each of  \n two sets of variables. We demonstrate the idea with a practical Wiberg  \n algorithm for L1 bundle adjustment\, the first algorithm for that problem. We  \n also show that one Wiberg minimization can be nested inside another\,  \n effectively removing two of three sets of variables from a minimization. We  \n demonstrate this idea with a nested Wiberg algorithm for L1 projective bundle  \n adjustment\, solving for camera matrices\, points\, and projective  \n depths.  Bio:  Dennis Strelow is an engineer on Google’s computer vision  \n research team.  He is interested in structure-from-motion\, vision-aided  \n navigation\, sensor fusion\, and image classification.  His recent work  \n includes Wiberg optimization\, image feature hashing for improved image  \n classification\, and fast local descriptors for large-scale image and video  \n classification.  Previously\, he worked at Quantapoint and Honeywell\, and  \n received his B.S.\, M.S.\, and Ph.D. degrees from the University of Wisconsin\,  \n University of Illinois\, and Carnegie Mellon\, respectively. 
END:VEVENT
BEGIN:VEVENT
UID:calendar.7942.field_date.0.15
SUMMARY:Mark Hill - Efficiently Enabling Conventional Block Sizes for Very Large  \n Die-stacked DRAM Caches
DTSTAMP:20130618T164412Z
DTSTART:20111129T220000Z
DTEND:20111129T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/mark-hill-efficiently-enabling-conventional-block-sizes-very-large-die-stacked-dram-caches
DESCRIPTION:Room: CS 1221 This is Micro 2011 practice talk for a paper co-authored with  \n Gabriel H. Loh of AMD. Die-stacking technology enables multiple layers of  \n DRAM to be integrated with multicore processors. A promising use of stacked  \n DRAM is as a cache\, since its capacity is insufficient to be all of main  \n memory (for all but some embedded systems). However\, a 1GB DRAM cache with  \n 64-byte blocks requires 96MB of tag storage. Placing these tags on-chip is  \n impractical (larger than on-chip L3s) while putting them in DRAM is slow (two  \n full DRAM accesses for tag and data). Larger blocks and sub-blocking are  \n possible\, but less robust due to fragmentation. This work efficiently enables  \n conventional block sizes for very large die-stacked DRAM caches with two  \n innovations. First\, we make hits faster than just storing tags in stacked  \n DRAM by scheduling the tag and data accesses as a compound access so the data  \n access is always a row buffer hit. Second\, we make misses faster with a  \n MissMap that eschews stacked-DRAM access on all misses. Like extreme  \n sub-blocking\, our implementation of the MissMap stores a vector of  \n block-valid bits for each “page” in the DRAM cache. Unlike conventional  \n sub-blocking\, the MissMap (a) points to many more pages than can be stored in  \n the DRAM cache (making the effects of fragmentation rare) and (b) does not  \n point to the “way” that holds a block (but defers to the off-chip tags).  \n For the evaluated large-footprint commercial workloads\, the pro- posed cache  \n organization delivers performance that is within 92.9% of the performance  \n benefit of an ideal 1GB DRAM cache with an impractical 96MB on-chip SRAM tag  \n array. Biography (since conference talks are usually given by grad students\,  \n not senior professors): Hill began graduate studies in 1981 at Berkeley with  \n David Patterson and Alan Smith. Fortunately\, he has finished. He worked on  \n the SPUR multiprocessor workstation and his thesis included developing the 3C  \n model of cache behavior (compulsory\, capacity\, and conflict misses) and  \n identifying when direct-mapped caches can out-perform set-associative ones  \n ("Big and Dumb is Better"). Hill subsequently joined Wisconsin as an  \n assistant professor.
END:VEVENT
BEGIN:VEVENT
UID:calendar.8056.field_date.0.16
SUMMARY:Exploratory Genomic Analysis and Modeling of Ovarian Carcinoma
DTSTAMP:20130618T164412Z
DTSTART:20111129T220000Z
DTEND:20111129T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/exploratory-genomic-analysis-and-modeling-ovarian-carcinoma
LOCATION:Biotechnology Center...
DESCRIPTION:Kevin Eng Department of Biostatistics and Medical Informatics UW-Madison  \n ---------------- Advanced-stage epithelial ovarian cancer is the leading  \n cause of death due to gynecological malignancy and\, unlike other cancers\, the  \n five-year survival rate has remained relatively constant for the last 30  \n years. This mortality is attributable to heterogeneous disease\, frequent  \n recurrence and acquired chemoresistance\, all of which complicate the attempt  \n to improve treatment outcomes and to generate models of the disease. In the  \n main\, I will discuss a number of hypotheses derived from our analysis of the  \n Cancer Genome Atlas' ovarian pilot project\, a multi-modal and high-throughput  \n molecular survey. Our focus on patterns of expression in frequently-mutated  \n "core" signaling pathways motivates statistical models that profile  \n individual disease\, correlate with response to drug therapies and promise  \n individual prognosis. Two shorter vignettes will discuss our early results  \n from studies of specific pathways/mechanisms and developing statistical  \n models of the clinical disease management process.
END:VEVENT
BEGIN:VEVENT
UID:calendar.8059.field_date.0.17
SUMMARY:staff lunch
DTSTAMP:20130618T164412Z
DTSTART:20111130T174500Z
DTEND:20111130T193000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/staff-lunch
LOCATION:5331
END:VEVENT
BEGIN:VEVENT
UID:calendar.8049.field_date.0.18
SUMMARY:Optimal Multi-dimensional Mechanisms via Multi- to Single-agent Reduction
DTSTAMP:20130618T164412Z
DTSTART:20111130T213000Z
DTEND:20111130T223000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/optimal-multi-dimensional-mechanisms-multi-single-agent-reduction
LOCATION:4310 CS
DESCRIPTION:Speaker: Nima Haghpanah\, Northwestern University We generalize the  \n ``marginal revenue'' approach to optimal mechanism design to environments  \n with multi-dimensional agent preferences. This approach can be viewed as a  \n reduction from multi- to single-agent mechanism design. For a single agent\,  \n marginal revenue describes the change in performance as a function of the ex  \n ante probability that the agent is served. When the distribution of  \n preferences is well behaved\, the optimal mechanism for multiple agents is to  \n serve the agents with the highest marginal revenue. This is a generalization  \n of Myerson's (1981) virtual-value-based construction. When the distribution  \n of preferences is not well behaved\, the marginal revenue approach fails\, and  \n so does any virtual-value-based construction. In these environments we  \n consider performance as a function of the interim allocation rule. Given a  \n construction of the performance-optimal mechanism for any interim allocation  \n rule for a single agent\, the optimal mechanism for multiple agents can be  \n constructed. The construction requires joint feasibility of the interim  \n allocation rule. This joint feasibility is known as Border's (1991)  \n condition. In general\, Border's condition has exponentially many constraints\;  \n however\, we give a polynomial time separation oracle from which the optimal  \n mechanism can be tractably constructed.
END:VEVENT
BEGIN:VEVENT
UID:calendar.7944.field_date.0.19
SUMMARY:MICRO Practice Talks
DTSTAMP:20130618T164412Z
DTSTART:20111130T220000Z
DTEND:20111130T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/micro-practice-talks
DESCRIPTION:Room: CS 1221Marc de Kruijf This work introduces a new processor  \n architecture\, the idempotent processor architecture\, that allows speculative  \n execution without the need for hardware checkpoints to recover from  \n mis-speculation. Idempotent processors execute programs as a sequence of  \n compiler-constructed idempotent (re-executable) regions\, and the nature of  \n these regions allows state to be reproduced by re-execution\, obviating the  \n need for hardware recovery support. The work builds upon the insight that  \n programs naturally decompose into a series of idempotent regions and that  \n these regions can be large. The talk will cover how idempotent processor  \n architecture simpli fies the design of in-order processors. Idempotent  \n processor architecture eliminates much of the complexities in modern in-order  \n processors by allowing instructions to retire out of order with support for  \n re-execution when necessary to recover precise state. Our quantitative  \n results show that we obtain a geometric mean performance increase of 4.4% (up  \n to 25% and beyond) while maintaining an overall reduction in power and  \n hardware complexity.  ************************************ Gagan Gupta:  \n Dataflow Execution of Sequential Imperative Programs on Multicore  \n Architectures As multicore processors become the default\, researchers are  \n aggressively looking for program execution models that make it easier to use  \n the available resources. Multithreaded programming models that rely on  \n statically-parallel programs have gained prevalence. Most of the existing  \n research is directed at adapting and enhancing such models\, alleviating their  \n drawbacks\, and simplifying their usage. This paper takes a different approach  \n and proposes a novel execution model to achieve parallel execution of  \n statically-sequential programs. It dynamically parallelizes the execution of  \n suitably-written sequential programs\, in a dataflow fashion\, on multiple  \n processing cores. Significantly\, the execution is race-free and determinate.  \n Thus the model eases program development and yet exploits available  \n parallelism. This presentation describes the implementation of a software  \n runtime library that implements the proposed execution model on existing  \n commercial multicore machines. We present results from experiments running  \n benchmark programs\, using both the proposed technique as well as traditional  \n parallel programming\, on three different systems. We find that in addition to  \n easing the development of the benchmarks\, the approach is resource-efficient  \n and achieves performance similar to the traditional approach\, using stock  \n compilers\, operating systems and hardware\, despite the overheads of an  \n all-software implementation of the model.
END:VEVENT
BEGIN:VEVENT
UID:calendar.8057.field_date.0.20
SUMMARY:Meet the AI Group: Profs. Rob Nowak and Ben Recht
DTSTAMP:20130618T164412Z
DTSTART:20111201T220000Z
DTEND:20111201T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/meet-ai-group-profs-rob-nowak-and-ben-recht
LOCATION:3310cs
DESCRIPTION:Come meet members of the UW Artificial Intelligence community! This is a  \n chance for new and returning students to get a feel for AI research being  \n done here at UW. It is also an opportunity to introduce students to  \n professors and researchers they may want to work with in the future. This  \n week with feature talks from Prof. Rob Nowak and Prof. Ben Recht Questions\,  \n contact bgibson@cs.wisc.edu Light refreshments will be served.
END:VEVENT
BEGIN:VEVENT
UID:calendar.8058.field_date.0.21
SUMMARY:Investigating Modern-day Disease by Using Metabolic Models of Bacteria from  \n the Past
DTSTAMP:20130618T164412Z
DTSTART:20111206T220000Z
DTEND:20111206T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/investigating-modern-day-disease-using-metabolic-models-bacteria-past
LOCATION:Biotechnology Center...
DESCRIPTION:David Baumler Genome Center of Wisconsin UW-Madison No other family of  \n microorganisms has had a greater impact on human health then the  \n Enterobacteriaceae\, and these bacteria have evolved into a wide variety of  \n commensal and human\, plant\, and avian pathogens. These organisms have  \n diverged from a common ancestor ~300-500 million years ago (MYA)\, and little  \n is known about ancestral metabolism. Using a paleo systems biology approach  \n the metabolism of ancient microorganisms has been investigated through  \n construction of metabolic models using either ancient genomic DNA (such as  \n the Yersinia pestis genome that has been recently sequenced from human  \n corpses that were victims of the 2nd pandemic of the black plague ~1300 A.D.)  \n or through a comparison of 72 enterobacterial genomes of modern descendents.  \n I will present an analysis of the ability of these ancient metabolic models  \n to utilize 300 substrates and how some of these metabolic strategies are used  \n in numerous human niche locations where modern-day enterobacteria cause  \n disease. Overall this work will demonstrate that models of ancient bacteria  \n can be used to accurately predict metabolism and to derive new targets to  \n control human disease.
END:VEVENT
BEGIN:VEVENT
UID:calendar.8079.field_date.0.22
SUMMARY:Meet the AI Group: Profs. Li Zhang and David Page
DTSTAMP:20130618T164412Z
DTSTART:20111208T220000Z
DTEND:20111208T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/meet-ai-group-profs-li-zhang-and-david-page
LOCATION:3310cs
DESCRIPTION:Come meet members of the UW Artificial Intelligence community! This is a  \n chance for new and returning students to get a feel for AI research being  \n done here at UW. It is also an opportunity to introduce students to  \n professors and researchers they may want to work with in the future. This  \n week with feature talks from Prof. Li Zhang and Prof. David Page Questions\,  \n contact bgibson@cs.wisc.edu Light refreshments will be served.
END:VEVENT
BEGIN:VEVENT
UID:calendar.8053.field_date.0.23
SUMMARY:Analysis of Software Artifacts – CS706 project presentations I
DTSTAMP:20130618T164412Z
DTSTART:20111209T170000Z
DTEND:20111209T175400Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/analysis-software-artifacts-%E2%80%93-cs706-project-presentations-i
LOCATION:1240 CS
DESCRIPTION:You are cordially invited to attend the class project presentations for  \n CS706: Analysis of Software Artifacts. Come learn about a variety of  \n innovative approaches to helping humans build software that works. Attend as  \n many or as few talks as you wish. 11:00am - 11:18am: Taming File Systems  \n Resource Leak 11:18am - 11:36am: Package Dependency Customization and  \n Visualization 11:36am - 11:54am: How to Diagnose Performance Bugs? Visit the  \n complete CS706 Project Presentation Schedule for more information\, including  \n talk abstracts. See also additional CS706 presentation sessions on Monday\,  \n December 12 and Wednesday\, December 14.
END:VEVENT
BEGIN:VEVENT
UID:calendar.7959.field_date.0.24
SUMMARY:Building knowledge bases from the web
DTSTAMP:20130618T164412Z
DTSTART:20111212T170000Z
DTEND:20111212T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/building-knowledge-bases-web
DESCRIPTION:Rajeev Rastogi\,   VP and Head of Yahoo! Labs BangaloreAbstract: The web is a  \n vast repository of human knowledge. Extracting structured data from web pages  \n can enable applications like comparison shopping\, and lead to improved  \n ranking and rendering of search results. In this talk\, I will describe two  \n efforts at Yahoo! Labs to extract records from pages at web scale. The first  \n is a wrapper induction system that handles end-to-end extraction tasks from  \n clustering web pages to learning XPath extraction rules to relearning rules  \n when sites change. The system has been deployed in production within Yahoo!  \n to extract more than 200 million records from ~200 web sites. The second  \n effort exploits content redundancy on the web to automatically extract  \n records without human supervision. Starting with a seed database\, we  \n determine values in the pages of each new site that match attribute values in  \n the seed records. We devise a new notion of similarity for matching  \n templatized attribute content\, and an apriori style algorithm that exploits  \n templatized page structure to prune spurious attribute matches.Room:  \n CS 2310 
END:VEVENT
BEGIN:VEVENT
UID:calendar.8054.field_date.0.25
SUMMARY:Analysis of Software Artifacts – CS706 project presentations II
DTSTAMP:20130618T164412Z
DTSTART:20111212T170000Z
DTEND:20111212T181200Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/analysis-software-artifacts-%E2%80%93-cs706-project-presentations-ii
LOCATION:1240 CS
DESCRIPTION:You are cordially invited to attend the class project presentations for  \n CS706: Analysis of Software Artifacts. Come learn about a variety of  \n innovative approaches to helping humans build software that works. Attend as  \n many or as few talks as you wish. 11:00am - 11:18am: Specializing Functions  \n for Sliced Executables 11:18am - 11:36am: Toward Better Tools for Debuggers  \n 11:36am - 11:54am: Fuzz Testing Google NaCl 11:54am - 12:12pm: MaliciOS: An  \n Environment for Advance Identification of Runtime Failures Visit the complete  \n CS706 Project Presentation Schedule for more information\, including talk  \n abstracts. See also additional CS706 presentation sessions on Friday\,  \n December 9 and Wednesday\, December 14.
END:VEVENT
BEGIN:VEVENT
UID:calendar.8055.field_date.0.26
SUMMARY:Analysis of Software Artifacts – CS706 project presentations III
DTSTAMP:20130618T164412Z
DTSTART:20111214T170000Z
DTEND:20111214T181200Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/analysis-software-artifacts-%E2%80%93-cs706-project-presentations-iii
LOCATION:1240 CS
DESCRIPTION:You are cordially invited to attend the class project presentations for  \n CS706: Analysis of Software Artifacts. Come learn about a variety of  \n innovative approaches to helping humans build software that works. Attend as  \n many or as few talks as you wish. 11:00am - 11:18am: Static Verification of  \n XPCOM Clients using Dehydra 11:18am - 11:36am: Semi-Supervised Statistical  \n Debugging of Non-Fatal Errors 11:36am - 11:54am: Overcoming Concolic Testing  \n Limitations 11:54am - 12:12pm: Accelerating Data Race Detection with Memory  \n Ranges Visit the complete CS706 Project Presentation Schedule for more  \n information\, including talk abstracts. See also additional CS706 presentation  \n sessions on Friday\, December 9 and Monday\, December 12.
END:VEVENT
BEGIN:VEVENT
UID:calendar.8097.field_date.0.27
SUMMARY:CS/Psych-770 HCI Class Poster Session
DTSTAMP:20130618T164412Z
DTSTART:20111216T213000Z
DTEND:20111216T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/cspsych-770-hci-class-poster-session
LOCATION:1st floor lobby
DESCRIPTION:All members of the department is invited to join a poster session with  \n projects from the HCI Class. Cookies and chocolate will be served. Below are  \n abstracts for the posters. Effects of facial expression in video- and  \n avatar-based computer-mediated communication Erdem Kaya\, Tomislav Pejsa\, &  \n Magdalena Rychlowska  The goal of this study is to explore collaborative  \n behavior and rapport in different modes of computer-mediated communication  \n (CMC). We tested whether the well-described in literature positive effects of  \n smile are present in video and 3D-avatar-based CMC. A between-subjects  \n laboratory experiment compared the effects of smiles shown in prerecorded and  \n rendered video sequences of a human and an avatar\, respectively\, displayed  \n during the iterated prisoner’s dilemma game. Self-reported and behavioral  \n measures included collaborative behavior of the participant\, likability and  \n rapport. Human game partners were perceived as more likable compared to  \n avatars. Smiles of the game partner predicted likability and rapport\, but not  \n collaboration. Importantly\, these effects were only significant for male  \n participants. Other effects of gender were observed: females reported higher  \n likability of both types of game partners\, smiled for longer time. Also\,  \n males were more collaborative with human compared to avatar game partners.    \n Comparing Administration Methods of Psychological Instruments Tony McDonald &  \n Lauren Meyer  Most psychological instruments were designed and normed with  \n paper and pencil administration\, but many researchers are now using computers  \n instead. Studies indicate there may be differences between these modes. How  \n does the mode of administration affect instruments’ psychometric  \n properties? How do participants perceive the experience? Twenty-seven  \n UW-Madison undergrad and graduate students completed four instruments (Social  \n Responsiveness Scale\, Interpersonal Trust Scale\, modified NASA Task Load  \n Index\, experience) in a 2 (mode: computer vs. paper) by 2 (location: lab vs.  \n home) between subjects design. Data from the computer conditions (n = 15)  \n were less reliable than from the paper conditions (n = 12)\, IPTS Cronbach’s  \n α = 0.39 vs. 0.76. Participants took less time in the computer conditions\, p  \n = 0.004\, and there was an interaction with location\, p = 0.01. Participants  \n in the computer conditions correctly judged the computer administration to be  \n faster\, p = 0.04. Participants prefer surveys on paper and in the lab\, p <  \n 0.001.   Do Only Humans Cheat? Exploring the Role of Cheating in a Video  \n Game J.J. De Simone\, Tessa Verbruggen\, & Li-Hsiang Kuo In sports and board  \n games\, an opponent cheating is typically met with disdain\, anger\, and  \n disengagement by the other players. However\, work has yet to address the role  \n of AI cheating in video games. Participants played either a cheating or  \n non-cheating version of a modified open source tower defense game. Results  \n indicate that when an AI competitor cheats\, players perceive the opponent as  \n being more human. Cheating also increases player aggravation\, but does not  \n impact presence and enjoyment of the experience. Game designers can integrate  \n subtle levels of cheating into AI opponents without any real negative  \n responses from the players. The data indicate that minor levels of cheating  \n might also increase player engagement with video games.   Effects of robot's  \n feedback under time pressure Heemoon Chae\, Mai Lee Chang\, & Nisha Kiran  \n During human-robot interaction under task time pressure\, a robot’s feedback  \n must be carefully designed since humans are under more stress and have less  \n tolerance of the robot’s mistakes. Thus far\, no research has investigated  \n the effects of a robot’s feedback under time pressure. Our study examines  \n the effects of a robot’s feedback on performance\, usefulness\, and  \n satisfaction under task time pressure. Participants were asked to find words  \n in a Word Search. The independent variables included feedback (none vs.  \n verbal) and time pressure (none vs. high). Feedback included both compliments  \n and hints. The dependent variables were usefulness\, satisfaction\, workload\,  \n and words per minute. The results show that the participants found the robot  \n to be useful\, assisting\, and raising alertness when the robot gave verbal  \n feedback.   The Effects of Text and Robotic Agents on Deception Detection  \n Wesley Miller & Michael Seaholm When people attempt to conceal the truth from  \n others\, they typically exhibit what are known as deception cues\, identifiable  \n indications that the person in question is speaking untruthfully. For this  \n study\, we were interested in seeing how well an individual can separate truth  \n from falsehood when a small subset of these deception cues are exhibited by  \n not only human\, but also robotic and text-based agents. Participants were  \n tasked with interacting with recordings of each agent in some predetermined  \n order\, asking them questions from a preset list. Participants then rate each  \n response in terms of perceived truthfulness. Once they have interacted with  \n all three agents\, participants were asked to rate the overall trustworthiness  \n of each agent. The results of our study indicate that participants were able  \n to detect deception more reliably from the human agent and that statements  \n given by the text-based agent were consistently ranked more truthful than the  \n statements of the other agents.   An Adaptive Autonomous System for Second  \n Language Acquisition Young-Bum Kim\, Pallavika Ramaswamy\, & Soyoun Kim Second  \n language acquisition poses a unique challenge in that it combines elements of  \n rote memory with skill integration and creative productions. To handle these  \n challenges\, we harness the power of frequent self-testing by designing a  \n Computer Aided Language Learning System(CALL) system. We explore whether  \n people can learn a second language\, by evaluating learner responses to a  \n language translation task\, with our CALL system. Our CALL system can  \n automatically assess the responses and also adapt the questions according to  \n the level of the learner. We comment about the efficiency of our adaptive  \n CALL system with a one-size-fits-all approach and see how it affects learning  \n curve of people with different knowledge bases . To this end\, we  show how  \n we can (1) estimate the semantic and syntactic difficulty of a sentence\,  \n (2)  autonomously grade the response\, and (3) select  problems according to  \n the assessed level of the learner.  In an uncontrolled online experiment\, we  \n asked the participants to translate English sentences into Korean sentences.  \n A third of the participants are given randomly chosen questions and    \n assessed manually\, and another third of the participants are given randomly  \n chosen questions but are assessed with our autonomous assessment system. A  \n final third of the participants are given adaptively chosen questions  \n according by our system. With this information\, we evaluate how the adaptive  \n autonomous assessment framework affected the participants' task performance.
END:VEVENT
BEGIN:VEVENT
UID:calendar.8067.field_date.0.28
SUMMARY:CS736 Project Poster Session
DTSTAMP:20130618T164412Z
DTSTART:20111219T160000Z
DTEND:20111219T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/cs736-project-poster-session
LOCATION:2310CS
DESCRIPTION:The CS736 (Advanced Operating Systems) class will present posters of their  \n final projects.  Come hear about this great collection of projects:Improved  \n APIs and Programming Models for Deterministic Multithreading\, Aarti Basant\,  \n James Paton\, Seth PollenAbstract: Dthreads is a deterministic multithreading  \n library which conforms to the pthreads API and programming model. However\,  \n the pthreads API was originally designed to handle nondeterministic  \n threading\, and using it to control a deterministic library masks the actual  \n semantics and introduces needless complexity. We propose a new\, simplified  \n API for deterministic multithreading\, consisting of just two calls which lock  \n and unlock a globally shared mutex. We show that this new API\, when combined  \n with the semantics of Dthreads\, is sufficiently powerful for general use and  \n can be used as the basis for more complex and powerful synchronization  \n operations\, such as condition variables and barriers.We then evaluate the new  \n API by incorporating its use into the PARSEC benchmark suite. We present  \n benchmark data to support the conclusion that our API changes do not  \n negatively impact performance. In addition\, we present a case study of a  \n PARSEC program that is broken by linking to the Dthreads library. We show  \n that a Dthreads-aware port of the program can fix the problem\, albeit at a  \n cost to performance. We conclude by outlining avenues for future  \n work. Race-Guided Random Testing for Multithreaded Applications\,  Dongdong  \n Deng\, Xiaoyang GaoAbstract: Exposing concurrency bugs in multithreaded  \n applications is difficult because of large interleaving space. Also\, buggy  \n interleaving may occur with extremely low probability\, making repeated  \n executions of an application inefficient to trigger concurrency bugs. With  \n our race detector\, we can focus on the instruction pairsdirectly related to  \n data races\, reducing the interleaving space significantly. To expose  \n concurrency bugs\, we suspend the execution of an application by inserting  \n random length of delay into randomly selected instructions. We analyzed the  \n causes of concurrency bugs as well as buggy interleaving in four  \n multithreaded applications and our experiment results show that we can  \n discover data races and expose concurrency bugs effectively in these  \n applications. The Pedigree of Linux\, Zev Weiss\, Matthew WolfeAbstract: We  \n have performed a source-level study of the variations between different  \n versions of the Linux kernel as packaged with six modern Linux  \n distributions.  Our goal was to provide a summary of the primary differences  \n that have emerged between operating systems all nominally referred to as  \n "Linux".We performed a detailed examination of the source code modifications  \n made in each of the respective kernels\, six in total.  We report on our  \n findings by presenting a broad sampling of new features added to the Linux  \n kernel by each distribution\, which are often quite recognizably relevant to  \n that particular distribution's intended use. Ultimately we find that there  \n are significant differences in the  feature sets offered by the various  \n versions of the kernel\, but do not see any evidence of divergence to an  \n extent that would break backward compatibility or cause compatibility  \n problems for portable programs (so long as they were not written specifically  \n to use a particular new  feature offered on a certain  \n distribution). Tree-Based Density Clustering using Graphics Processors\, Evan  \n Samanas\, Benjamin WeltonAbstract: Tree-Based Overlay Networks (TBONs) have  \n proven to be a powerful and scalable communication model for building tools  \n for massively parallel and distributed systems\, but this model is less proven  \n for building applications. We present as a case study our software that uses  \n the TBON model to efficiently execute a popular clustering algorithm usually  \n run on a single node while utilizing Graphics Processing Units (GPUs) in a  \n TBON. Our non-GPU implementation provides up to a 23x speedup at 32 nodes  \n over the algorithm running on a single node\, while a non-TBON based solution  \n is able to achieve 20x speedup before declining. A 7x-23x speedup is also  \n seen in using a single GPU to run the same clustering algorithm. We then give  \n an example use of clustering at large-scale by attempting to track the  \n movement and outbreak of influenza in the United States. Implementing a Data  \n race detector with the Dyninst API\, Xiaozhu Meng\, Luke Pillar\, Nairan  \n ZhangAbstract: Data races are difficult to manually debug\, so a tool that can  \n detect Data races automatically is quite useful to writing multi-threaded  \n programs. In this paper\, we describe the implementation of a Data race  \n detector using the  DyninstAPI\, an efficient binary-instrumentation tool.  \n Our Data race detector implements a modification of the Vector Clock  \n algorithm\, based on the  “happened-before” partial-ordering  \n relationship. We implement this algorithm for the pthread library\,  \n instrumenting the pthread synchronization primitives. In our tests\, our  \n detector reports data races successfully. We conduct a comprehensive study on  \n the performance and scalability of our detector and  give an analysis of  \n performance bottlenecks. We believe that our detector has  the potential to  \n be used on real programs after some further optimization. A procfs Interface  \n for Debugging in Linux\, Srinivas Govindan\, Sathya GunasekarAbstract: Many  \n Unix based operating systems provide support for debugging through their  \n procfs interface. Linux however\, provides a debugging support through the  \n ptrace system call\, while at the same time it exposes a process through its  \n procfs interface to some extent. We extend the current procfs interface in  \n Linux to support all the debugging features provided by ptrace and also  \n provide additional features enabling a debugger to perform better. Debuggers  \n can read/write to specific files in /proc to perform the  debugging tasks.  \n Our debugging interface can read/write to the process address space\,  \n registers and supports tracing specific system calls that are explicitly of  \n interest to the debugger. We also enable the debugger to choose not to be  \n notified about specific signals delivered to a traced process that it is  \n not interested in. Our simple debugging interface performs as good as ptrace  \n for normal debugging and can perform better while tracing system calls. It  \n also provides much faster reads and writes to the address space of a process  \n compared to ptrace which only allows access one word at a time. Debugging  \n support through /proc file system in Linux\, Victor Bittorf\, Igor  \n CanadiAbstract: Using the virtual file system in Linux\, we implemented  \n debugger primitives through /proc file system. Our new interface is intended  \n as replacement for the existing ptrace system call. We discuss differences  \n and advantages of our model. We created a proctrace library to  \n semi-transparently emulate ptrace using our interface. While we cannot  \n perfectly emulate ptrace\, we did have some success in porting strace. We also  \n created our own tracer\, itrace\, which can be run simultaneously with  \n gdb. Reverse Engineering of Data Structures using Dyninst\, Eric Lederer\,  \n Srividya DossAbstract: The knowledge of a program’s data structures along  \n with their syntactic and semantic definitions can be highly valuable in a  \n variety of forensic and security applications. In this paper\, we attempt to  \n infer rich information regarding the types of these data structures from only  \n the binary executable of a program. We first instrument the binary with the  \n the Dyninst API such that we can monitor data flow across memory locations  \n and registers when the instrumented binary is executed.  At runtime\, each  \n memory access is tracked by a shadow array\, and we construct a graph of  \n dependencies between tracked memory locations.  We use the REWARDS [1]  \n dynamic analysis technique to automatically reverse engineer data structures  \n from this information by propagating type information across these  \n dependencies between memory locations.  The key to type resolution is the  \n execution of a type-revealing instruction called a “type sink\,” typically  \n system calls and standard library functions with well-defined parameter and  \n return type information.  We have a developed a prototype tool\, Pumpkin\,  \n which outputs a visualization of the data structures present in the  \n binary’s memory address space and the layout of data structures within the  \n binary’s custom function stack frames. Refreshments will be served.
END:VEVENT
BEGIN:VEVENT
UID:calendar.8098.field_date.0.29
SUMMARY:Automated Concurrency-Bug Diagnosis and Fixing
DTSTAMP:20130618T164412Z
DTSTART:20111220T190000Z
DTEND:20111220T210000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/automated-concurrency-bug-diagnosis-and-fixing
LOCATION:CS4310
DESCRIPTION:  Student: Guoliang Jin Committee: Shan Lu (Advisor)  \n                  Ben Liblit                   \n Michael Swift                  Andrea Arpaci-Dusseau 
END:VEVENT
BEGIN:VEVENT
UID:calendar.8071.field_date.0.30
SUMMARY:CS758 project presentations
DTSTAMP:20130618T164412Z
DTSTART:20111221T160000Z
DTEND:20111221T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/cs758-project-presentations
LOCATION:3310CS
END:VEVENT
BEGIN:VEVENT
UID:calendar.8096.field_date.0.31
SUMMARY:Redundancy elimination as a network primitive
DTSTAMP:20130618T164412Z
DTSTART:20111221T160000Z
DTEND:20111221T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/redundancy-elimination-network-primitive
LOCATION:CS 2310
DESCRIPTION:Student: Ashok Anand  Committee:       Prof. Aditya Akella (Advisor)    \n                            Prof. Remzi-Arpaci Dusseau        \n                        Prof. Suman Banerjee                \n                Prof. Parmesh Ramanathan                      \n          Prof. Srinivasan Seshan    
END:VEVENT
BEGIN:VEVENT
UID:calendar.8112.field_date.0.32
SUMMARY:Towards an Inductive Game Theory: Extracting Strategies and Causal Networks  \n from Time-series Data
DTSTAMP:20130618T164412Z
DTSTART:20120123T220000Z
DTEND:20120123T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/towards-inductive-game-theory-extracting-strategies-and-causal-networks-time-series-data
LOCATION:WID room 3280B (Anyone...
DESCRIPTION:Speaker: Jessica Flack Co-Director of the Center for Complexity & Collective  \n Computation\, Wisconsin Institute for Discovery University of  \n Wisconsin-Madison   Presentation Title: Towards an Inductive Game Theory:  \n Extracting Strategies and Causal Networks from Time-series Data   Abstract:  \n One weakness of game theory is that the strategies individuals play and the  \n payoffs associated with those strategies are posited rather than derived from  \n data. I will discuss an approach my collaborators and I have been developing  \n for extracting strategies\, or decision-making rules\, from social event time  \n series data. In social groups the collective implementation of the strategies  \n by multiple individuals produces functionally important statistical features  \n of social structure\, like the distribution of power or distribution of fight  \n sizes. Hence it would be useful to have a meso-scale description of how  \n alternative aggregate social properties arise from different combinations of  \n strategies at the individual level. This meso-scale description can be  \n formalized in terms of causal networks the topology of which specifies how  \n the strategies (and individuals) are combined. I will discuss how these  \n causal networks can be constructed from strategy data.    
END:VEVENT
BEGIN:VEVENT
UID:calendar.8936.field_date.0.33
SUMMARY:Computational Mixed Integer Nonlinear Programming
DTSTAMP:20130618T164412Z
DTSTART:20120130T220000Z
DTEND:20120130T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/computational-mixed-integer-nonlinear-programming
LOCATION:WID room 3280 (teaching...
DESCRIPTION:Coffee at 3:30pm in WID room 3280 and Presentation at 4:00pm Anyone without  \n WID access can use the special event elevator on the WID 1st floor (near ALDO  \n Café) to access the 3rd floor teaching lab   Speaker: Jeffrey Linderoth  \n Professor\, Industrial and Systems Engineering\, UW-Madison Presentation Title:  \n Computational Mixed Integer Nonlinear Programming   Abstract: Optimization  \n models that contain both nonlinear and discrete components are known as Mixed  \n Integer Nonlinear Programs (MINLP)s. MINLPs have been proposed for many  \n diverse and important applications. Unfortunately\, current algorithms and  \n software are often unable to solve practically-sized instances of these  \n important models. This talk will give an overview of our ongoing research  \n efforts aimed at addressing the mismatch between natural MINLP models and  \n robust algorithms and software for solving them.  An overarching theme of  \n our research is to study and exploit model-specific mathematical  \n structures.  The talk will discuss solution techniques for both convex and  \n nonconvex MINLPs.            
END:VEVENT
BEGIN:VEVENT
UID:calendar.8963.field_date.0.34
SUMMARY:The Interaction of Islet Amyloid Polypeptide with Phospholipid Bilayers and  \n Micelles
DTSTAMP:20130618T164412Z
DTSTART:20120131T220000Z
DTEND:20120131T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/interaction-islet-amyloid-polypeptide-phospholipid-bilayers-and-micelles
LOCATION:Biotechnology Center...
DESCRIPTION:Computational Biology Kyle Hoffman Department of Chemical and Biological  \n Engineering UW-Madison 1/31/12 The Interaction of Islet Amyloid Polypeptide  \n with Phospholipid Bilayers and Micelles Tuesday 4:00 pm Biotechnology Center  \n Auditorium\, 425 Henry Mall Abstract: The aggregation of the 37 residue  \n polypeptide human amylin is strongly linked to the development of Type II  \n Diabetes. The human and rat sequences of the peptide differ by only 6  \n residues\; however\, the human version aggregates into fibrils while the rat  \n version does not. While the conformations these proteins adopt in solution  \n have been studied extensively\, their structures in the presence of lipid  \n bilayers where the aggregation process is substantially accelerated are  \n unknown. By studying the differences between the structures\, insight into the  \n aggregation process can be gained. Using molecular dynamics and other  \n advanced sampling techniques\, we show that the structure of human and rat  \n amylin near a lipid bilayer are a random coil. However\, in the presence of a  \n micelle\, the rat amylin adopts a more α-helical conformation. These  \n structures will allow us to begin studying the aggregation process in the  \n presence of membranes.
END:VEVENT
BEGIN:VEVENT
UID:calendar.8964.field_date.0.35
SUMMARY:Dark Silicon and the End of Multicore Scaling.
DTSTAMP:20130618T164412Z
DTSTART:20120131T220000Z
DTEND:20120131T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/dark-silicon-and-end-multicore-scaling
LOCATION:CS 1221
DESCRIPTION:Speaker: Emily Blem Abstract: Since 2005\, processor designers have increased  \n core counts to exploit Moore’s Law scaling\, rather than focusing on  \n single-core performance. The failure of Dennard scaling\, to which the shift  \n to multicore parts is partially a response\, may soon limit multicore  \n scaling just as single-core scaling has been curtailed. This talk will cover  \n models used to examine multicore scaling limits by combining device scaling\,  \n single-core scaling\, and multicore scaling to measure the speedup potential  \n for a set of parallel workloads for the next five technology  \n generations.  The multicore designs we study include single threaded  \n CPU-like and massively threaded GPU-like multicore chip organizations with  \n symmetric\, asymmetric\, dynamic\, and composed topologies. The study shows  \n that regardless of chip organization and topology\, multicore scaling is  \n power limited to a degree not widely appreciated by the computing  \n community.    Bio: I am a Computer Science PhD student at University of  \n Wisconsin-Madison working with Professor Karu Sankaralingam.  My research  \n interests include future-looking computer architectures\, the application of  \n analytic modeling to performance and power projections\, and the impacts of  \n transistor scaling and new technologies.
END:VEVENT
BEGIN:VEVENT
UID:calendar.8937.field_date.0.36
SUMMARY:Confidence regions for stochastic variational inequalities.
DTSTAMP:20130618T164412Z
DTSTART:20120206T220000Z
DTEND:20120206T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/confidence-regions-stochastic-variational-inequalities
LOCATION:WID room 3280 (teaching...
DESCRIPTION:Coffee served at 3:30pm in WID room 3280\, Presentation at 4:00pm Anyone  \n without WID access can use the special events elevator on WID 1st floor (near  \n Aldo Café) to access the 3rd floor teaching lab.   Speaker: Shu Lu\,  \n Assistant Professor\, Statistics & Operation Research\, University of North  \n Carolina\, Chapel Hill\, NC  Presentation Title: Confidence regions for  \n stochastic variational inequalities.   Abstract: The sample average  \n approximation (SAA) method is a basic approach for solving stochastic  \n variational inequalities (SVI). It is well known that under appropriate  \n conditions the SAA solutions provide asymptotically consistent point  \n estimators for the true solution to an SVI. We propose a method to build  \n asymptotically exact confidence regions for the true solution that are  \n computable from the SAA solutions\, by exploiting the precise geometric  \n structure of the variational inequalities and by appealing to certain large  \n deviations probability estimates. We justify this method theoretically by  \n establishing a precise limit theorem\, and apply it to complementarity  \n problems. The material of this presentation is based on joint work with  \n Amarjit Budhiraja.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.8985.field_date.0.37
SUMMARY:Persistent Homology and its Application to Brain Network Modeling
DTSTAMP:20130618T164412Z
DTSTART:20120207T220000Z
DTEND:20120207T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/persistent-homology-and-its-application-brain-network-modeling
LOCATION:Biotechnology Center...
DESCRIPTION:Speaker: Moo K. Chung\, PhD Associate Professor Department of Biostatistics  \n and Medical Informatics\; Waisman Laboratory for Brain Imaging and Behavior  \n University of Wisconsin-Madison Tuesday\, February 7th 4:00 p.m. Biotechnology  \n Center Auditorium 425 Henry Mall PERSISTENT HOMOLOGY AND ITS APPLICATION TO  \n BRAIN NETWORK MODELING Abstract: By borrowing heavily from persistent  \n homology\, topological data analysis\, tries to infer high dimensional  \n structure by understanding how local neighborhoods of the structure connect  \n to each other to form a topology. A simplical complex with a scale parameter  \n usually represents the local connectivity. Then instead of looking at the  \n topology at a fixed scale\, the persistent homology observes the changes of  \n topology over the varying scales and finds the most persistent topological  \n features\, which are robust under perturbation. The persistent homology has  \n been previously used as a form of topological feature reduction in  \n quantifying amount of the gray matter in the magnetic resonance images (MRI)  \n of the human brain. Recently we have succeeded in modeling the brain network  \n by introducing the concept of graph filtration\, which decomposes an arbitrary  \n weighted graph uniquely into a sequence of unweighted graphs. The prosed  \n method is applied in characterizing autistic brain network using MRI\, PET and  \n DTI. We will also present our recent attempt at extending the framework to  \n scale spaces induced by heat diffusion. Brief Bio: Moo K. Chung\, Ph.D.  \n (http://www.stat.wisc.edu/~mchung)\, is an Associate Professor in the  \n Department of Biostatistics and Medical Informatics at the University of  \n Wisconsin-Madison. Also affiliated with the Waisman Laboratory for Brain  \n Imaging and Behavior. He is also affiliated with the Department of Brain and  \n Cognitive Sciences\, Seoul National University since 2009. Dr. Chung received  \n Ph.D. in statistics from McGill University under Keith J. Worsley and James  \n O. Ramsay in 2001. Dr. Chung¹s main research area is computational  \n neuroanatomy\, where noninvasive brain imaging modalities such as magnetic  \n resonance imaging (MRI) and diffusion tensor imaging (DTI) are used to map  \n the spatiotemporal dynamics of the human brain. In 2011\, Dr. Chung received  \n the Editor's Award for best paper published in Journal of Speech\, Language\,  \n and Hearing Research\, where CT images of vocal tract structures are analyzed.  \n Recent research focus has been on the topological data analysis using  \n persistent homology and its application to brain imaging and networks. He is  \n currently writing a 400-page research monograph on computational neuroanatomy  \n that will be published this year.
END:VEVENT
BEGIN:VEVENT
UID:calendar.8990.field_date.0.38
SUMMARY:The DySER Architecture: Continuing the Faster\, Smaller\, Greener Paradigm of  \n Microprocessors
DTSTAMP:20130618T164412Z
DTSTART:20120207T220000Z
DTEND:20120207T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/dyser-architecture-continuing-faster-smaller-greener-paradigm-microprocessors
LOCATION:CS 1221
DESCRIPTION:Speaker: Karu Sankaralingam   Abstract: The era of faster\, smaller\, greener  \n (more power efficient) transistors in every successive generation appears to  \n be dead. Instead\, processor architects and microarchitects are going to be  \n partially burdened with power-efficiently and energy-efficiently improving  \n performance with technology scaling providing density improvements "alone".  \n In this talk\, I will describe the DySER project which investigates ways for  \n dynamically specializing datapaths to energy-efficiently improve performance.  \n DySER attempts to provide a truly general purpose accelerator\, avoiding  \n radical changes to software development\, ISA\, or microarchitecture. Our key  \n insights are the following. First\, applications execute in phases and these  \n phases can be determined by creating a path-tree of basic-blocks rooted at  \n the inner-most loop. Second\, specialized datapaths corresponding to these  \n path-trees\, can be constructed by interconnecting a set of heterogeneous  \n computation units with a circuit-switched network. On simulation-based  \n results for the SPEC and Parsec benchmarks\, the DySER architecture provides  \n geometric mean speedup of 2.1X (1.15X to 10X)\, and geometric mean energy  \n reduction of 40% (up to 70%). On highly tuned throughput kernels\, DySER  \n outperforms hand-optimized SSE code. To demonstrate the architecture is  \n practical\, we have completed a prototype FPGA implementation of DySER  \n integrated with the OpenSPARC processor. This design runs unmodified Linux  \n and runs actual applications generated by the DySER compiler. The talk will  \n provide the main insights behind the DySER architecture\, quantitative  \n results\, and outline challenges and possibilities in using it in a modern  \n microprocessor and particularly as an accelerator in a mobile processor.
END:VEVENT
BEGIN:VEVENT
UID:calendar.8129.field_date.0.39
SUMMARY:How good are simple auctions?
DTSTAMP:20130618T164412Z
DTSTART:20120209T220000Z
DTEND:20120209T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/how-good-are-simple-auctions
LOCATION:CS 1240
DESCRIPTION:Speaker: Eva Tardos Affiliation: Cornell University Abstract: Auctions for  \n selling advertisement space have been the main mechanism used to monetize  \n Internet services. These auctions give rise to a number of interesting  \n challenges not traditionally considered by auction theory. Search  \n advertising gives rise of an enormous number of auctions running  \n simultaneously\, necessitating the use of extremely simple mechanisms.  \n Traditional auction theory tells us how to design optimal auctions  \n (maximizing welfare or revenue)\, but typically results in designs that are  \n too complex for the web environment\, where it is essential for auctions to  \n have extremely clear and simple design. Over the last 10+ years we have  \n developed a good understanding of many games naturally arising in the  \n context of Internet or web services from the perspective of the resulting  \n social welfare\, including a good understanding of games modeling selfish  \n traffic routing\, service location\, bandwidth sharing among others. In this  \n talk we will consider auctions from this perspective in various settings  \n including the commonly used auction format Generalized Second Price.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.8130.field_date.0.40
SUMMARY:Sequential Auctions and the Curse of Simultaneity
DTSTAMP:20130618T164412Z
DTSTART:20120210T203000Z
DTEND:20120210T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/tbd-0
LOCATION:CS 4310
DESCRIPTION:Speaker: Eva Tardos Affiliation: Cornell University Abstract: Typical models  \n of strategic interactions in computer science use simultaneous move  \n games. However\, in applications simultaneity is often hard or impossible to  \n achieve. In this talk we study the robustness of the Nash Equilibrium when  \n the assumption of simultaneity is dropped. In some classes of games  \n sequential action significantly improves the quality of the predicted  \n solution\, resulting in much more natural and better quality prediction. In  \n the context of auctions\, sequential implementation gives players interesting  \n strategic opportunities and introduces externalities in the auction game.  \n  We study such games in an attempt to understand the role of simultaneity  \n in simple auction formats (such as simultaneous first or second price  \n auctions).  
END:VEVENT
BEGIN:VEVENT
UID:calendar.8938.field_date.0.41
SUMMARY:Topological optimization
DTSTAMP:20130618T164412Z
DTSTART:20120213T220000Z
DTEND:20120213T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/topological-optimization
LOCATION:WID room 3280 (teaching...
DESCRIPTION:Coffee served at 3:30pm in WID room 3280\, Presentation starts at 4:00pm    \n Anyone without WID access can use the special event elevator on the WID 1st  \n floor (near ALDO Café) to access the 3rd floor teaching lab   Speaker:  \n Jean-Luc Thiffeault Associate Professor\, Department of Mathematics\,  \n UW-Madison   Presentation Title: Topological optimization   Abstract:  \n Topological chaos is a type of chaotic behavior that is 'forced' by the  \n motion of obstacles in some domain.  I will review two topological  \n approaches\, with applications in particular to stirring and mixing in fluid  \n dynamics.  The first approach involves constructing systems such that the  \n fluid motion is topologically complex\, usually by imposing a specific motion  \n of rods.  I will then discuss optimization strategies that can be  \n implemented.  The second approach is diagnostic\, where flow characteristics  \n are deduced from observations of periodic or random orbits and their  \n topological properties.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.8999.field_date.0.42
SUMMARY:Using Multi-relational Data and Machine Learning to Improve Breast Cancer  \n Diagnosis
DTSTAMP:20130618T164412Z
DTSTART:20120214T220000Z
DTEND:20120214T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/using-multi-relational-data-and-machine-learning-improve-breast-cancer-diagnosis
LOCATION:Biotechnology Center...
DESCRIPTION:Speaker: Elizabeth S. Burnside\, MD PhD Department of Radiology\; Department of  \n Biostatistics and Medical Informatics University of Wisconsin School of  \n Medicine and Public Health Using Multi-relational Data and Machine Learning  \n to Improve Breast Cancer Diagnosis Tuesday 4:00 pm Biotechnology Center  \n Auditorium\, 425 Henry Mall Abstract:     In the new era of "-omic"-based  \n research\, many scientists have shifted from the study of the individual parts  \n of a system to the system itself. This new paradigm focuses on a  \n comprehensive collection of a fundamental data type that can provide a  \n platform for a myriad of research directions on a given level ranging from  \n the subcellular to the population. However\, developing methodologies that  \n integrate these rich data sources to inform and improve healthcare decisions  \n on the patient level remains challenging. Our team of physicians\, computer  \n scientists\, and industrial engineers at the University of Wisconsin has  \n collaborated for the last decade to develop methods to improve breast cancer  \n diagnostic decision-making using inductive logic programming\, statistical  \n relational learning\, advice-based-learning\, and other cutting-edge machine  \n learning techniques. Our algorithms are designed to utilize the  \n ever-expanding\, multi-relational data that predicts breast cancer including:  \n genetic\, imaging\, and epidemiologic risk factors. This talk will present an  \n overview of our research programs and provide a vision of the future of  \n computational methods in the domain of breast cancer risk prediction.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9000.field_date.0.43
SUMMARY:Acoherent Shared Memory
DTSTAMP:20130618T164412Z
DTSTART:20120214T220000Z
DTEND:20120214T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/acoherent-shared-memory
LOCATION:CS 1221
DESCRIPTION:Speaker: Derek Hower   Abstract: Given the current trends in computing\, it  \n may be a good time to rethink coherent memory in multicores. The extremes of  \n computing where we find multicores\, like mobile devices and datacenters\, have  \n drastically different characteristics than the discrete multiprocessors where  \n coherence emerged. Also\, as architects pursue deeper integration of  \n accelerators like GPUs on die\, full on-chip coherence may no longer be  \n optimal. In addition to rethinking coherence because of ecosystem changes\, we  \n show that coherence prevents a key opportunity for software to exploit  \n on-chip caches. Software runtimes that require isolation (e.g.\, transactional  \n memory\, database logs) usually make copies of data that they modify -- copies  \n which\, ironically\, may already exist in cache. Because coherence hides caches  \n from software\, applications are forced to make expensive and redundant  \n copies\, and consequently often require hardware acceleration for good  \n performance. In this talk\, we propose Acoherent Shared Memory (ASM)\, a new  \n abstraction that gives software control over how hardware manages memory. ASM  \n introduces the concept of acoherent memory that is neither coherent (it  \n allows multiple versions of the same address) nor incoherent (hardware  \n resolves differences when it matters). Acoherence exposes private memory to  \n threads\, and uses a checkout/checkin abstract to manage data movement between  \n private and shared memory. We show that ASM performs comparably to coherent  \n systems for existing workloads and can accelerate software data isolation up  \n to 49%\, paving the way for fast software-only runtimes such as transactional  \n memory.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9003.field_date.0.44
SUMMARY:Learning networks in biology: opportunities and challenges
DTSTAMP:20130618T164412Z
DTSTART:20120220T220000Z
DTEND:20120220T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/learning-networks-biology-opportunities-and-challenges
LOCATION:Wisconsin Institute for...
DESCRIPTION:Coffee served at 3:30 in conjunction with WID Discovey.  Non-building  \n occupants meet at room 3280. Sushmita Roy\, Asst Professor\, Biostatistics and  \n Medical Informatics\, UW-Madison Title: Learning networks in biology:  \n opportunities and challenges A systems-level understanding of how living  \n system function requires us to identify the parts of a system and the  \n interactions between these parts. More importantly\, the interactions of a  \n system may be “condition-specific”\, where a condition could represent a  \n different environmental stress\, a cell-type\, a disease or different  \n organisms. Advances in biotechnology are enabling us to identify the parts of  \n a system\, however\, identifying the interactions remains a difficult problem.  \n I will discuss the inference and analysis of these networks in the usual  \n setting of learning a single network\, and then in a more interesting setting  \n of simultaneous learning of multiple networks. I will present the challenges  \n in learning these networks\, specifically\, when we aim to do this at the  \n genome-scale\, with thousands of nodes. I will present some of our approaches\,  \n based on probabilistic graphical models\, to address this problem that allows  \n us to use search heuristics and incorporate prior knowledge into system to  \n make the problem more tractable and the networks more biologically realistic.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9035.field_date.0.45
SUMMARY:A case for abstract yet accurate cortical models
DTSTAMP:20130618T164412Z
DTSTART:20120221T220000Z
DTEND:20120221T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/case-abstract-yet-accurate-cortical-models
LOCATION:CS 1221
DESCRIPTION:Speaker:  Atif Hashmi   Abstract: Developing abstract models to emulate the  \n functionality of low-level complex mechanisms has contributed significantly  \n towards several advancements in computer architecture. These abstract models  \n are relatively easy to implement\, they are computationally efficient\, and  \n they allow the architects to study interactions among high-level modules  \n while avoiding the intricacies and implementation details of low-level  \n constructs. A similar approach has been adopted in neural network design\;  \n however\, within this paradigm the high-level abstract models do not fully  \n capture the key aspects of their biological inspirations. As a result\, these  \n high-level neural abstractions significantly compromise the power of the  \n underlying biological architecture. This talk describes cortical columns as  \n one of the high-level biologically inspired computational abstractions. It  \n provides insights into the biological basis of the column model and  \n demonstrates the advantages of such a neural abstraction. Afterwards\, it  \n introduces spiking neural networks as an interesting low-level neural  \n abstraction and describes various powerful mechanisms provided by spiking  \n neurons that are not captured by the aforementioned column model. This  \n disconnection between the high and low level neural models severely limits  \n the applicability of high-level neural models towards performing complex  \n tasks like perception\, attention\, and decision-making. High-level neural  \n models have their advantages\; however\, to properly develop such models\, their  \n low-level biological inspirations must be studied and modeled in details.  \n Only then can truly intelligent autonomous systems be developed. Computer  \n architects are uniquely positioned for this task because\, unlike biologists\,  \n their goal is to steer this research towards useful computing systems and  \n applications. Although both the elementary components and the resulting  \n biologically-inspired systems are quite different from existing computational  \n hardware\, similar architectural approaches can and should be used to model  \n these biological abstractions. Biography: Dr. Hashmi obtained his M.S. and  \n Ph.D. in Electrical Engineering from the University of Wisconsin (UW) –  \n Madison. His research focuses on understanding structural and functional  \n properties of the mammalian brain and developing computational models  \n inspired by these properties. Presently\, Dr. Hashmi is a post-doctoral  \n research associate in the Sleep and Consciousness labs at UW – Madison  \n working with Professor Giulio Tononi and is studying complex neuronal  \n dynamics and their role in perception\, attention\, and decision-making. He is  \n also part of the UW – Madison research team collaborating with IBM on the  \n SyNAPSE project.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9032.field_date.0.46
SUMMARY:Quantitative Transcription and Replication Dynamics of Vesicular Stomatitis  \n Virus – Towards a Predictive Model of Viral Growth
DTSTAMP:20130618T164412Z
DTSTART:20120221T220000Z
DTEND:20120221T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/quantitative-transcription-and-replication-dynamics-vesicular-stomatitis-virus-%E2%80%93-towards-predi
LOCATION:Biotechnology Center...
DESCRIPTION:Speaker: Collin M. Timm Department of Chemical and Biological Engineering  \n University of Wisconsin-Madison 2/21/12 Quantitative Transcription and  \n Replication Dynamics of Vesicular Stomatitis Virus – Towards a Predictive  \n Model of Viral Growth Tuesday 4:00 pm Biotechnology Center Auditorium\, 425  \n Henry Mall Abstract: Viruses are a major threat to human health causing  \n diseases ranging from the common cold to many cancers. Anti-viral treatments  \n are currently designed by targeting specific viral mechanisms but due to many  \n interconnected events the effect on the overall infection cycle is difficult  \n to predict. Using vesicular stomatitis virus as a model virus we are making  \n quantitative measurements of viral RNA to determine rates of viral  \n transcription and replication. These rates are used to inform a mechanistic  \n model of virus growth to predict viral behavior in new environments\, such as  \n drug treated cells. Vesicular stomatitis virus (VSV) is an RNA virus which  \n means the genes are encoded on an RNA strand. After the viral genome is  \n delivered to the cell\, primary transcription begins. As viral mRNA are  \n translated into viral protein the viral polymerase switches from a  \n transcription mode to a replication mode\, and genome replication begins.  \n Further accumulation of viral proteins leads to packaging of new virus  \n particles\, and the infection can proceed in subsequent cells.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9016.field_date.0.47
SUMMARY:AI Seminar/AIRG: Burr Settles
DTSTAMP:20130618T164412Z
DTSTART:20120222T220000Z
DTEND:20120222T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/ai-seminarairg-burr-settles
LOCATION:1240 CS
DESCRIPTION:  AISEM in collaboration with AIRG welcomes returning alumnus Burr Settles.  \n TITLE Interactive Machine Learning: Combining Learning Strategies with Humans  \n in the Loop ABSTRACT People learn by interacting with their teachers.  Why  \n not machines?  What would it take to develop software that can learn how to  \n solve problems by interacting and collaborating with humans?  This talk will  \n describe my efforts to develop such systems\, with the goal of training  \n effective machine learners more quickly and economically.  In particular\, I  \n focus on two projects in natural language processing that combine multiple  \n learning strategies: incorporating domain knowledge (taking advice in the  \n form of human-provided rules)\, active learning (asking "questions" of human  \n annotators)\, and semi-supervised learning (attempting to "teach itself" by  \n extrapolating what has been learned onto abundant\, unlabeled data).  \n  Empirical results from user experiments show that these approaches are  \n superior to their state-of-the-art "passive" learning counterparts.  \n  Interestingly\, these experiments provide initial insights into human  \n "teaching" behavior as well\, suggesting ways in which human factors can and  \n should be taken into account.  I will also briefly discuss opportunities for  \n interactive learning in other areas\, such as supporting online communities\,  \n creative work\, and biological discovery. SHORT BIO Burr Settles is a  \n Postdoctoral Fellow in the Machine Learning Department at Carnegie Mellon  \n University. He received a PhD in Computer Sciences from the University of  \n Wisconsin-Madison in 2008\, with additional studies in Linguistics and  \n Biology. His current research focuses on interactive machine learning that  \n resembles a "dialogue" of decision-making and knowledge acquisition between  \n computers and humans\, with applications in natural language processing\,  \n biology\, and social computing. He recently organized workshops at the ICML  \n and NAACL conferences on these topics\, and is the author of a popular  \n literature survey on active learning (active-learning.net). He also runs the  \n website FAWM.ORG\, prefers sandals to shoes\, and plays guitar in the  \n Pittsburgh pop band Delicious Pastries.
END:VEVENT
BEGIN:VEVENT
UID:calendar.8971.field_date.0.48
SUMMARY:HCI/AI Joint Seminar: Andrea Thomaz
DTSTAMP:20130618T164412Z
DTSTART:20120223T220000Z
DTEND:20120223T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/hciai-joint-seminar-andrea-thomaz
LOCATION:1240 CS
DESCRIPTION:Designing Learning Interactions for Robots Speaker: Andrea Thomaz\, Georgia  \n Institute of Technology Abstract: In this talk I present recent work from the  \n Socially Intelligent Machines Lab at Georgia Tech. One of the focuses of our  \n lab is on Socially Guided Machine Learning\, building robot systems that can  \n learn from everyday human teachers. We look at standard Machine Learning  \n interactions and redesign interfaces and algorithms to support the collection  \n of learning input from naive humans. This talk covers results on high-level  \n task goal learning\, low-level skill learning\, and active learning  \n interactions using several humanoid robot platforms. Bio: Andrea L. Thomaz is  \n an Assistant Professor of Interactive Computing at the Georgia Institute of  \n Technology. She directs the Socially Intelligent Machines lab\, which is  \n affiliated with the Robotics and Intelligent Machines (RIM) Center and with  \n the Graphics Visualization and Usability (GVU) Center. She earned a B.S. in  \n Electrical and Computer Engineering from the University of Texas at Austin in  \n 1999\, and Sc.M. and Ph.D. degrees from MIT in 2002 and 2006. Dr. Thomaz is  \n published in the areas of Artificial Intelligence\, Robotics\, Human-Robot  \n Interaction\, and Human-Computer Interaction. She received an ONR Young  \n Investigator Award in 2008\, and an NSF CAREER award in 2010. Her work has  \n been featured on the front page of the New York Times\, and in 2009 she was  \n named one of MIT TR35.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9030.field_date.0.49
SUMMARY:Online Matching with Concave Returns
DTSTAMP:20130618T164412Z
DTSTART:20120224T203000Z
DTEND:20120224T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/online-matching-concave-returns
LOCATION:4310 CS
DESCRIPTION:Speaker: Nikhil Devanur Affiliation: Microsoft Research(Redmond) Abstract: We  \n consider a generalization of the Adwords problem by allowing arbitrary  \n concave returns\, and we characterize the optimal competitive ratio achievable  \n for this problem. The problem considers a sequence of items arriving online  \n that have to be allocated to agents\, with different agents bidding different  \n amounts. The objective function is the sum\, over each agent $i$\, a  \n monotonically non-decreasing concave function $M_i : \Real_+ \rightarrow  \n \Real_+$ of the total amount allocated to $i$. All variants of online  \n matching problems (including the Adwords problem) studied in the literature  \n consider the special case of budgeted linear functions\, that is\, functions of  \n the form $M_i( u_i) = \min \{u_i\,B_i\}$ for some constant $B_i$. The  \n distinguishing feature here is in allowing arbitrary concave returns. The  \n main result of this paper is that for each concave function $M$\, there exists  \n a constant $F(M) \leq 1$ such that - there exists an algorithm with  \n competitive ratio of $\min_i\{ F(M_i) \}$\, independent of the sequence of  \n items. - No algorithm has a competitive ratio larger than $F(M)$ over all  \n instances with $M_i= M$ for all $i$. Our algorithm is based on the  \n primal-dual paradigm and makes use of convex programming duality. The upper  \n bounds are obtained by formulating the task of finding the right  \n counterexample as an optimization problem. This path takes us through the  \n calculus of variations which deals with optimizing over continuous functions.  \n The algorithm and the upper bound are related to each other via a set of  \n differential equations\, which points to a certain kind of duality between  \n them. Joint work with Kamal Jain
END:VEVENT
BEGIN:VEVENT
UID:calendar.9025.field_date.0.50
SUMMARY:The Interplay of Convexity and Algorithmic Algebra in Optimization and  \n Systems Analysis
DTSTAMP:20130618T164412Z
DTSTART:20120227T220000Z
DTEND:20120227T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/interplay-convexity-and-algorithmic-algebra-optimization-and-systems-analysis
LOCATION:Engineering Hall\, room...
DESCRIPTION:Speaker: Amir Ali Ahmadi\, Postdoctoral Associate (Candidate for WID  \n Optimization Faculty Position) Lab for Information & Decision Systems (LIDS)  \n and Computer Science & Artificial Intelligence Laboratory\, Massachusetts  \n Institute of Technology           Abstract: Exciting recent developments  \n at the interface of computational algebra and convex optimization have led to  \n major algorithmic advances in a broad range of problems in optimization and  \n systems theory. I will start this talk by giving an overview of these  \n techniques and presenting applications in continuous and combinatorial  \n optimization\, statistics\, and control theory. I will then focus on two recent  \n results on computational and algebraic aspects of convexity in optimization:  \n (i) I will show that deciding convexity of polynomials of degree as low as  \n four is strongly NP-hard. This solves a problem that appeared as one of seven  \n open problems in complexity theory for numerical optimization in 1992. (ii) I  \n will introduce an algebraic\, semi-definite programming (SDP) based sufficient  \n condition for convexity known as sum-of-squares convexity and present a  \n complete characterization of the cases where it is equivalent to convexity.  \n This characterization draws an interesting parallel to a seminal1888 result  \n of Hilbert in real algebraic geometry. In the final part of the talk\, I will  \n move on to a problem with numerous applications in engineering and sciences:  \n understanding the asymptotic behavior of linear dynamical systems under  \n uncertainty. I will tackle this problem with a novel class of SDP-based  \n algorithms (with provable guarantees) that are based on new connections  \n between ideas from control theory and the theory of finite automata.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9002.field_date.0.51
SUMMARY:Computer Architecture Seminar: FabScalar: Composing Synthesizable RTL Designs  \n of Arbitrary Cores within a Canonical Superscalar Template
DTSTAMP:20130618T164412Z
DTSTART:20120228T220000Z
DTEND:20120228T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/computer-architecture-seminar-fabscalar-composing-synthesizable-rtl-designs-arbitrary-cores-wi
LOCATION:CS 1221
DESCRIPTION:Speaker: Jayneel Gandhi Abstract: A growing body of work has compiled a  \n strong case for the single-ISA heterogeneous multi-core paradigm. A  \n single-ISA heterogeneous multi-core provides multiple\, differently-designed  \n superscalar core types that can streamline the execution of diverse programs  \n and program phases. No prior research has addressed the “Achilles’  \n heel” of this paradigm: design and verification effort is multiplied by the  \n number of different core types. This work frames superscalar processors in a  \n canonical form\, so that it becomes feasible to quickly design many cores that  \n differ in the three major superscalar dimensions: superscalar width\, pipeline  \n depth\, and sizes of structures for extracting instruction- level parallelism  \n (ILP). From this idea\, we develop a toolset\, called FabScalar\, for  \n automatically composing the synthesizable register-transfer-level (RTL)  \n designs of arbitrary cores within a canonical superscalar template. The  \n template defines canonical pipeline stages and interfaces among them. A  \n Canonical Pipeline Stage Library (CPSL) provides many implementations of each  \n canonical pipeline stage\, that differ in their superscalar width and depth of  \n sub-pipelining. An RTL generation tool uses the template and CPSL to  \n automatically generate an overall core of desired configuration. Validation  \n experiments are performed along three fronts to evaluate the quality of RTL  \n designs generated by FabScalar: functional and performance  \n (instructions-per-cycle (IPC)) validation\, timing validation (cycle time)\,  \n and confirmation of suitability for standard ASIC flows. With FabScalar\, a  \n chip with many different superscalar core types is conceivable.     Bio:  \n Jayneel Gandhi is currently a Graduate Student in ECE department pursuing PhD  \n under Dr. Mark Hill. He received his bachelor’s degree in Information and  \n Communication Technology (ICT) from Dhirubhai Ambani Institute of Information  \n and Communication Technology (DA-IICT)\, Gandhinagar\, India in 2008. Jayneel  \n Gandhi received his Master's degree in Computer Engineering at North Carolina  \n State Univeristy (NCSU) under the guidance of Dr. Eric Rotenberg in 2010. He  \n has gained industrial experience during internships with ST Microelectronics  \n and most recently with Intel. He research interests lies in the area of  \n Computer Architecture and VLSI Design.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9045.field_date.0.52
SUMMARY:A MRI Study of Sustained Neuromodulation Induced by Electrical Tongue  \n Stimulation in Balance Disorders
DTSTAMP:20130618T164412Z
DTSTART:20120228T220000Z
DTEND:20120228T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/mri-study-sustained-neuromodulation-induced-electrical-tongue-stimulation-balance-disorders
LOCATION:Biotechnology Center...
DESCRIPTION:Speaker: M. Elizabeth Meyerand\, PhD Chair and Professor\, Department of  \n Biomedical Engineering\; Professor\, Department of Medical Physics University  \n of Wisconsin-Madison Tuesday\, February 28th 4:00 p.m. Biotechnology Center  \n Auditorium 425 Henry Mall A MRI STUDY OF SUSTAINED NEUROMODULATION INDUCED BY  \n ELECTRICAL TONGUE STIMULATION IN BALANCE DISORDERS Research Interests /  \n Applied Neuro MRI Lab: http://www.neurofmri.bme.wisc.edu/ Our research is in  \n the field of magnetic resonance imaging (MRI) of the human brain. Our goal is  \n the development of new MR methods to visualize the structure and function of  \n the brain and to translate these methods to the hospital for clinical  \n diagnosis. One of the areas upon which we concentrate our research is  \n functional MRI (fMRI. FMRI allows us to visualize both the temporal and  \n spatial patterns of brain activity in response to different stimuli. We are  \n particularly interested in the development of new analysis methods to improve  \n our understanding of brain function. In addition to analyzing brain  \n activation\, we are also developing techniques to explore brain connectivity  \n using diffusion tensor imaging (DTI) and the concept of effective  \n connectivity. As implemented in MRI\, DTI is a noninvasive imaging technique  \n that can be used to probe the intrinsic diffusion characteristics of tissue.  \n Techniques for diffusion imaging are evolving rapidly. Diffusion MRI research  \n has been shown to have important applications\, especially in stroke\, the  \n effects of tumors\, degenerative diseases and brain injury. Effective  \n connectivity describes the integration within and between functionally  \n specialized areas of the brain. Regions of the brain are located using fMRI.  \n Integration of these regions is achieved through the information gained from  \n DTI. We explore the effective connectivity in a variety of large-scale  \n neurocognitive networks using structural equation modeling.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9056.field_date.0.53
SUMMARY:Deborah Chasman Preliminary Exam
DTSTAMP:20130618T164412Z
DTSTART:20120229T180000Z
DTEND:20120229T200000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/deborah-chasman-preliminary-exam
LOCATION:3310 Computer Sciences
DESCRIPTION:Title:Inferring Host-Virus Interaction Networks Committee: Mark Craven  \n (advisor)\, Paul Ahlquist\, Michael Ferris\, David Page\, Xiaojin Zhu
END:VEVENT
BEGIN:VEVENT
UID:calendar.9040.field_date.0.54
SUMMARY:HCI Seminar: Interacting with and through agentic objects
DTSTAMP:20130618T164412Z
DTSTART:20120229T220000Z
DTEND:20120229T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/hci-seminar-interacting-and-through-agentic-objects
LOCATION:1240 CS
DESCRIPTION:HCI Seminar: Interacting with and through agentic objects Speaker: Leila  \n Takayama\, Willow Garage Inc. Abstract: We encounter and interact with  \n non-human agents every day when withdrawing cash from ATMs\, driving cars  \n with anti-lock brakes\, and tuning our thermostats. Through controlled  \n experiments and field studies\, this talk will examine the ways that people  \n make sense of agentic objects\, including (1) how we interact with agentic  \n objects like voice agents and personal robots\, and (2) how we  \n interact through agentic objects like telepresence robots. Drawing from  \n the theories that informed ubiquitous computing\, we can see how people make  \n sense of agentic objects\, thereby providing implications for both theory and  \n the design of interactive systems. Bio: Leila Takayama is a research  \n scientist at Willow Garage\, studying human-computer interaction and  \n human-robot interaction. Dr. Takayama completed her PhD in Communication at  \n Stanford University in 2008. She also holds a PhD minor in Psychology from  \n Stanford\, MA in Communication from Stanford\, and BAs in Psychology and  \n Cognitive Science from UC Berkeley (2003). During her graduate studies\, she  \n was a research assistant in the User Interface Research (UIR) group at Palo  \n Alto Research Center (PARC). Her thesis\, titled “Throwing Voices:  \n Investigating the Psychological Effects of the Spatial Location of Projected  \n Voices\,” won the Nathan Maccoby outstanding dissertation award. She is a  \n member of the global agenda council on robotics and smart devices for the  \n World Economic Forum as well as editor for the inaugural issue of the Journal  \n of Human-Robot Interaction.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9031.field_date.0.55
SUMMARY:Learning\, Inference\, and Control for Sustainable Energy
DTSTAMP:20130618T164412Z
DTSTART:20120301T220000Z
DTEND:20120301T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/learning-inference-and-control-sustainable-energy
LOCATION:1240 CS
DESCRIPTION:J. Zico Kolter Stanford University Ph.D. MIT Postdoc Sustainable energy  \n issues pose one of the largest challenges facing society: 84% of the world's  \n energy currently comes from fossil fuels\, raising major issues with climate  \n change\, energy security\, and the long-term availability of these sources.  \n Although energy domains span a huge range of different areas\, a common theme  \n in many modern energy tasks is the availability of large amounts of data\, and  \n the need to learn models\, make inferences\, and control the system based upon  \n this data. These are problems that require new methods in machine learning\,  \n probabilistic inference\, and control\, and where such algorithms can have a  \n profound impact on the energy space. In this talk I will look at two  \n particular tasks spanning different extremes of energy consumption and  \n generation and show how new algorithmic methods can play a pivotal role in  \n each. First\, on the energy consumption side\, I will present new techniques  \n for energy disaggregation\, the task of taking an aggregate power signal and  \n decomposing it into separate devices. This ability helps us understand how  \n energy is consumed in a building\, and studies have shown that just presenting  \n this information to users can directly lead to large energy savings. Unlike  \n previous approaches to this problem\, my work considers models that look  \n jointly at the entire signal and exploit the rich temporal structure of the  \n data. The key technical challenge here is the task of making inferences in  \n these high-dimensional\, factorized\, temporal models\, and I will present new  \n algorithms I have developed\, based upon convex relaxations of inference\, that  \n greatly outperform existing approaches on this task. Second\, on the energy  \n generation side\, I will present work on maximizing power output for wind  \n turbines in low-wind conditions. In particular\, I will present a novel policy  \n learning approach\, based upon trust-region optimization\, which is able to  \n maximize power using much less data than existing learning techniques\, and  \n which produces 30% more power than a purely model-based approach on an  \n experimental wind turbine.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9037.field_date.0.56
SUMMARY:Aleksandr “Sasha’ Aravkin\, Postdoctoral Fellow: Robust Statistical  \n Modeling for Geophysical Imaging and Kalman Smoothing
DTSTAMP:20130618T164412Z
DTSTART:20120305T180000Z
DTEND:20120305T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/robust-statistical-modeling-geophysical-imaging-and-kalman-smoothing
LOCATION:Wisconsin Institute for...
DESCRIPTION:Speaker: Aleksandr “Sasha’ Aravkin Postdoctoral Fellow\, Earth & Ocean  \n Science and Computer Science\, University of British Columbia\, Vancouver\, B.C.  \n   Presentation Title: Robust Statistical Modeling for Geophysical Imaging  \n and Kalman Smoothing   Abstract: For many inverse problems\, accuracy of data  \n is required by standard methods\, yet rarely achievable in practice. In many  \n applications\, data may contain large artifacts\, such as outliers caused by  \n measurement errors\, or physical phenomena not explained by the predictive  \n model. In this setting\, robust methods\, i.e. methods that find reasonable  \n results even in the face of gross errors\, are an appealing alternative  \n to pre-processing\, outlier removal\, or very complex modeling.   In this  \n talk\, we will discuss two applications: geophysical imaging and inference for  \n dynamical systems. In both cases\, we will show how to design robust methods  \n by modifying the statistical error models. We can then get robust solutions  \n by finding the maximum likelihood estimates for parameters in these modified  \n models. In order to solve these problems quickly\, optimization techniques  \n must exploit the underlying problem structure. We will highlight this  \n structure for both classes of applications\, and present numerical results to  \n show how the methods work in practice.   
END:VEVENT
BEGIN:VEVENT
UID:calendar.9014.field_date.0.57
SUMMARY:Algorithms\, Graph Theory\, and the Solution of Laplacian Linear Equations
DTSTAMP:20130618T164412Z
DTSTART:20120305T220000Z
DTEND:20120305T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/algorithms-graph-theory-and-solution-laplacian-linear-equations
LOCATION:1240
DESCRIPTION:Speaker: Dan Spielman Affiliation: Yale University Abstract: We survey  \n several fascinating concepts and algorithms in graph theory that arise in the  \n design of fast algorithms for solving linear equations in the Laplacian  \n matrices of graphs. We will begin by explaining why linear equations in these  \n matrices are so interesting. The problem of solving linear equations in these  \n matrices motivates a new notion of what it means for one graph to approximate  \n another. This leads to the problem of approximating graphs by sparse graphs.  \n Our algorithms for solving Laplacian linear equations will exploit  \n surprisingly strong approximations of graphs by sparse graphs\, and even by  \n trees. We will survey the roles that spectral graph theory\, random matrix  \n theory\, graph sparsification\, low-stretch spanning trees and local clustering  \n algorithms play in the design of fast algorithms for solving Laplacian linear  \n equations.   Bio: Daniel Alan Spielman received his B.A. in Mathematics and  \n Computer Science from Yale in 1992\, and his Ph.D in Applied Mathematics from  \n M.I.T. in 1995. He spent a year as a NSF Mathematical Sciences Postdoc in the  \n Computer Science Department at U.C. Berkeley\, and then taught in the Applied  \n Mathematics Department at M.I.T. until 2005. Since 2006\, he has been a  \n Professor of Applied Mathematics and Computer Science at Yale University. The  \n awards he has received include the 1995 ACM Doctoral Dissertation Award\, the  \n 2002 IEEE Information Theory Paper Award\, the 2008 Godel Prize\, the 2009  \n Fulkerson Prize\, and the 2010 Nevanlinna Prize. He is a fellow of the  \n Association for Computing Machinery. His main research interests include the  \n design and analysis of algorithms\, graph theory\, machine learning\,  \n error-correcting codes and combinatorial scientific   
END:VEVENT
BEGIN:VEVENT
UID:calendar.9057.field_date.0.58
SUMMARY:AI In the Wild - Monitoring the Earth from above: Mining satellite data to  \n map urbanization at local to global scales
DTSTAMP:20130618T164412Z
DTSTART:20120306T190000Z
DTEND:20120306T200000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/ai-wild-monitoring-earth-above-mining-satellite-data-map-urbanization-local-global-scales
LOCATION:1240cs
DESCRIPTION:  AISEM Presents Monitoring the Earth from above: Mining satellite data to  \n map urbanization at local to global scales Annemarie Schneider Assistant  \n Professor\, Nelson Institute’s Center for Sustainability and the Global  \n Environment and Department of Geography For the first time in history\, more  \n than 50 percent of the Earth’s population now live in cities\, towns and  \n settlements. From an environmental standpoint\, cities consume enormous  \n amounts of resources\, the by-products of urban activity and land use are  \n numerous\, and recent studies demonstrate that the ecological footprint of  \n many cities is significant and not sustainable. Cities are also emerging as  \n an important source of uncertainty in local\, regional and global-scale  \n biogeophysical processes. Urban land use influences local climates through  \n urban heat islands\, impervious surfaces alter sensible and latent heat  \n fluxes\, and recent evidence has suggested that cities may also significantly  \n affect precipitation regimes. Accurate and timely information on the  \n distribution and nature of urban areas is therefore critical to a wide array  \n of research questions related to the effect of humans on the regional and  \n global environment This talk will describe how remotely sensed data - used in  \n conjunction with data mining algorithms\, spatial statistics\, and  \n socio-economic and demographic data - have begun to play a substantive role  \n in investigating alterations of the Earth’s surface within and near  \n urbanized areas\, and in monitoring and modeling these changes. At the global  \n scale\, my recent work has focused on a new map of global urban extent  \n developed using the fusion of multiple remote sensing and ancillary data  \n inputs. At the local scale\, my research has aimed to understand urban land  \n use trajectories across multiple time points by quantifying rates and  \n patterns of expansion in a sample of 40 cities across the globe\, with a  \n specific focus on rapidly urbanizing regions in China. Finally\, I will review  \n the opportunities and challenges of using satellite data to monitor landscape  \n processes\, including the benefits of incorporating new processing algorithms  \n and unique data sources to characterize heterogeneous urban environments.  \n Bio: Annemarie Schneider in an Assistant Professor in the Nelson Institute  \n for Environmental Studies with an affiliated appointment in the Geography  \n Department. Dr. Schneider earned her M.A. and Ph.D. in Geography and  \n Environmental Science at Boston University. Before arriving in Madison in  \n 2007\, she was a faculty member at the Department of Geography and Institute  \n for Computational Earth System Science at the University of California\, Santa  \n Barbara.  This talk is part of the AI In the Wild series.  For more  \n information\, including upcoming talks\, click here.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9066.field_date.0.59
SUMMARY:Spectral Sparsification of Graphs and Approximations of Matrices
DTSTAMP:20130618T164412Z
DTSTART:20120306T203000Z
DTEND:20120306T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/spectral-sparsification-graphs-and-approximations-matrices
LOCATION:CS 3310
DESCRIPTION:Speaker: Daniel Spielman Affiliation: Yale University   Expander graphs can  \n be viewed as sparse approximations of complete graphs\, with Ramanujan  \n expanders providing the best possible approximation.   We formalize this  \n notion of approximation and ask how well a given graph can be approximated by  \n a sparse graph. We prove that every graph can be approximated by a sparse  \n graph almost as well as the complete graphs are approximated by the Ramanujan  \n expanders: our approximations employ at most twice as many edges to achieve  \n the same approximation factor.   We also present an efficient randomized  \n algorithm for constructing sparse approximations that only uses a logarithmic  \n factor more edges than optimal.   Our algorithms follow from the solution of  \n a problem in linear algebra. Given any expression of a rank-n symmetric  \n matrix A as a sum of rank-1 symmetric matrices\, we show that A can be well  \n approximated by a weighted sum of only O(n) of those rank-1 matrices.   This  \n is joint work with Joshua Batson\, Nikhil Srivastava and Shang-Hua Teng.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9034.field_date.0.60
SUMMARY:Address Translation in Virtualized Systems
DTSTAMP:20130618T164412Z
DTSTART:20120306T220000Z
DTEND:20120306T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/address-translation-virtualized-systems-0
LOCATION:CS 1221
DESCRIPTION:Speaker:  Lena Olson   Abstract:  Guests in a virtualized environment  \n expect a contiguous\, isolated physical address space. To provide this\, the  \n hypervisor translates guest physical addresses into host physical addresses\,  \n much like virtual addresses are translated into physical addresses in an  \n operating system. There are two approaches to managing these translations.  \n With shadow page tables\, the hypervisor traps on guest page table  \n modifications and updates the shadow page table\, a list of GVA->HPA  \n translations. This requires no hardware support but involves many costly VM  \n exits. Nested page tables are an alternative where both the guest and host  \n page tables are hardware walked. This eliminates the necessity of trapping  \n into the hypervisor when the guest page table is updated\, but for n-level  \n page tables can result in O(n*2) accesses to translate each address. I will  \n present a survey of recent work to speed up address translation\, including 2D  \n translation caching\, addition of a nested TLB\, and using a hashed page table  \n for the GPA->HPA translation.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9041.field_date.0.61
SUMMARY:Address Translation in Virtualized Systems
DTSTAMP:20130618T164412Z
DTSTART:20120306T220000Z
DTEND:20120306T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/address-translation-virtualized-systems
LOCATION:CS 1221
DESCRIPTION:Speaker: Lena Olson   Abstract:  Guests in a virtualized environment  \n expect a contiguous\, isolated physical address space. To provide this\, the  \n hypervisor translates guest physical addresses into host physical addresses\,  \n much like virtual addresses are translated into physical addresses in an  \n operating system. There are two approaches to managing these translations.  \n With shadow page tables\, the hypervisor traps on guest page table  \n modifications and updates the shadow page table\, a list of GVA->HPA  \n translations. This requires no hardware support but involves many costly VM  \n exits. Nested page tables are an alternative where both the guest and host  \n page tables are hardware walked. This eliminates the necessity of trapping  \n into the hypervisor when the guest page table is updated\, but for n-level  \n page tables can result in O(n*2) accesses to translate each address. I will  \n present a survey of recent work to speed up address translation\, including 2D  \n translation caching\, addition of a nested TLB\, and using a hashed page table  \n for the GPA->HPA translation.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9065.field_date.0.62
SUMMARY:Intelligent Personal Health Records
DTSTAMP:20130618T164412Z
DTSTART:20120306T220000Z
DTEND:20120306T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/intelligent-personal-health-records
LOCATION:Biotechnology Center...
DESCRIPTION:Speaker: Gang Luo IBM T.J. Watson Research Center 3/6/12 Intelligent Personal  \n Health Records Tuesday 4:00 pm Biotechnology Center Auditorium\, 425 Henry  \n Mall Abstract: Web-based personal health records (PHRs) are under massive  \n deployment. To improve PHR’s capability and usability and provide users  \n with personalized healthcare information to facilitate their daily activities  \n of living\, we proposed the concept of intelligent PHR (iPHR). iPHR introduces  \n and extends expert system technology\, Web search technology\, database trigger  \n technology\, and natural language generation technology into the PHR domain.  \n By extensively using medical knowledge and nursing knowledge\, iPHR can  \n anticipate users’ needs\, guide users to provide the most important  \n information about their medical condition\, and automatically form medical  \n queries. Our iPHR system currently provides four functions: questionnaire  \n guided search for disease information\, recommendation of home nursing  \n activities\, recommendation of home medical products\, and continuous user  \n monitoring. Upon the detection of an abnormal event that may have potential  \n medical impact\, the fourth function follows the push model of information  \n distribution and actively pushes related personalized healthcare information  \n to the user. This talk will present both the high-level design and some  \n implementation details of iPHR.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9081.field_date.0.63
SUMMARY:Mirek Riedewald : Scolopax: Supporting Exploratory Analysis of Scientific  \n Data
DTSTAMP:20130618T164412Z
DTSTART:20120312T160000Z
DTEND:20120312T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/scolopax-supporting-exploratory-analysis-scientific-data
LOCATION:CS 2310
DESCRIPTION:  Abstract: As the amount and complexity of data in many fields is rapidly  \n increasing\, new approaches are needed for exploratory analysis and scientific  \n discovery. Our Scolopax system's goal is to address these challenges with  \n novel  techniques for large-scale parallel data management. In this talk\, we  \n will present an overview of Scolopax and then focus on parallel processing of  \n joins. Joins combine information across data sets\, e.g.\, to discover  \n correlations. Our proposed join model simplifies reasoning about how to  \n assign computation tasks to processors in MapReduce and other parallel  \n environments. Using this model\, we derive a surprisingly simple randomized  \n algorithm\, called 1-Bucket-Random\, for implementing arbitrary theta-joins in  \n a single MapReduce job. This algorithm only requires minimal statistics  \n (input cardinality) and we provide proofs and strong evidence that for a  \n variety of join problems\, its latency is either close to optimal or the best  \n realizable option. For some popular joins we show how to improve over  \n 1-Bucket-Random by exploiting additional input  statistics. Various aspects  \n of Scolopax were published at premier data management and data mining venues  \n like SIGMOD\, VLDB\, ICDE\, ICML\, and ICDM. Bio: Mirek Riedewald received a  \n Ph.D. in computer science from the University of California at Santa Barbara  \n in 2002. After spending some time as a researcher at Cornell University and  \n as a visiting researcher at Microsoft Research\, he is now an Associate  \n Professor at Northeastern University. Dr. Riedewald's research interests are  \n in databases and data mining\, with an emphasis on designing scalable  \n techniques for data-driven science. Currently Dr. Riedewald is developing  \n novel approaches for parallel data processing and for mining observational  \n data. He has a track record of successful collaborations with scientists from  \n different domains\, including ornithology\, physics\, mechanical and aerospace  \n engineering\, and astronomy. His work has been published in the premier  \n peer-reviewed data management research venues like ACM SIGMOD\, VLDB\, IEEE  \n ICDE\, and IEEE TKDE\, as well as in domain science journals.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9083.field_date.0.64
SUMMARY:Bhavesh Mehta and Vikram Joshi: Getting a Million IOPS Through Code You Don't  \n Own
DTSTAMP:20130618T164412Z
DTSTART:20120312T170000Z
DTEND:20120312T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/getting-million-iops-through-code-you-dont-own
LOCATION:2310 CS
DESCRIPTION:ABSTRACT Virtualization has brought in disruptive change to the way we use  \n compute servers allowing for greater consolidation ratios\, reduction of  \n capital expenses and energy costs\, and ease of IT management. But along with  \n the benefits have come some serious challenges\, the biggest one being the I/O  \n bottleneck for virtual machines. Most hypervisors require the use of shared  \n storage. While the I/O demands of any single VM may not be great\, the  \n aggregate I/O requirements of many VMs running on a single server quickly add  \n up. Primary storage vendors have been the beneficiaries and monetized this  \n surprise blessing that landed on their laps. Flash memory has changed the way  \n we can look at both the I/O as well as the virtual memory subsystem of  \n operating systems\, allowing us to use Flash as a part of the new memory to  \n storage hierarchy. The ability to use Flash transparently in compute servers  \n to offload hundreds of thousands of IOPS from primary storage is now  \n theoretically possible but still a daunting task given the idiosyncratic  \n nature of the Flash medium and the need to seamlessly and transparently  \n modify commercial/proprietary operating systems to deal with high I/O loads.  \n This talk gets into the inner workings of unifying the memory I/O subsystems  \n of OS-es to open up the flood gates  BIO // Vikram Joshi Vikram is a VP &  \n Chief Technologist at Fusion-io. His technical background spans multiple  \n disciplines such as operating systems\, parallel and distributed systems\,  \n databases\, storage\, media and computer graphics. Prior to co-founding IO  \n Turbine (acquired by Fusion-io)\, he founded PixBlitz Studios which developed  \n high-definition virtual advertising technology for broadcast sports and  \n entertainment. At Oracle\, his work included doubling database performance on  \n 12-64 way SMPs and laying the foundation for the Exadata appliance group.  \n Vikram pioneered high-speed texture-mapped graphics for video\, worked  \n on video on demand\, and video game server (CosmoSoft) at Silicon Graphics.  \n At Sun Microsystems\, he worked on the Solaris virtual memory subsystem  \n to increase scalable OS performance up to 10X\, and on the  \n Spring Microkernel (Sunlabs). He holds a MS (Hons.) in Physics and a  \n BE (Hons.) in Engineering from the Birla Institute of Technology  \n and Science\, Pilani\, India. BIO // Bhavesh Mehta Bhavesh Mehta is a software  \n engineer at Fusion-io\, working in kernel group solving interesting problems  \n spanning storage and virtualization. He received his Master's from the  \n University of Wisconsin in 2005 and worked in the Multifacet research group  \n as a graduate student. Prior to joining Fusion-io he was a software engineer  \n in the Hypervisor Group at VMware\, where he contributed to wide range of  \n features like monitor\, vmkernel\, EFI\, fault-tolerence  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9061.field_date.0.65
SUMMARY:Optimal Power Flow and Demand Response
DTSTAMP:20130618T164412Z
DTSTART:20120312T203000Z
DTEND:20120312T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/optimal-power-flow-and-demand-response
LOCATION:Wisconsin Institute for...
DESCRIPTION:Speaker: Steven H. Low <?xml:namespace prefix = o /> Professor\, Computer  \n Science and Electrical Engineering\, California Institute of Technology\,  \n Pasadena\, CA Presentation Title: Optimal Power Flow and Demand Response  \n Optimal power flow (OPF) problems determine the most efficient power  \n generations\, reactive powers for voltage support\, or demand response. They  \n are well-known to be nonconvex and hence NP hard. In the first part of the  \n talk\, we propose relaxations that can be solved efficiently\, focusing on  \n radial networks.  Recently a sufficient condition is proved for general mesh  \n network under which a semidefinite relaxation is exact. We prove that\, if the  \n network is radial (tree)\, then the sufficient condition is always satisfied  \n and hence the semidefinite relaxation is always exact\, provided the  \n constraints on power flows satisfy a simple pattern. Using the branch flow  \n model for radial networks\, we propose a simple SOCP relaxation to OPF\, and  \n prove that it is exact. We apply this result to control voltage and reactive  \n power in distribution networks\, and present results from realistic simulation  \n of a Southern California distribution circuit. In the second part of the  \n talk\, we describe a simple model that integrates two-period electricity  \n markets\, uncertainty in renewable generation\, and real-time dynamic demand  \n response. A load serving entity decides its day-ahead procurement to optimize  \n expected social welfare a day before energy delivery. At delivery time when  \n renewable generation is realized\, it coordinates with users\, in a  \n decentralized manner\, to manage load and purchase real-time balancing power  \n in the real-time market\, if necessary. We derive the optimal day-ahead  \n decision\, propose real-time demand response algorithm\, and study the effect  \n of volume and variability of renewable generation on the optimal social  \n welfare.   (Joint work with Subhomesh Bose\, Mani Chandy\, Masoud Farivar\,  \n Lingwen Gan\, Dennice Gayme\, Libin Jiang\, Javad Lavaei\, Caltech\, and Chris  \n Clarke\, SCE.)    
END:VEVENT
BEGIN:VEVENT
UID:calendar.9036.field_date.0.66
SUMMARY:Motors\, Voters\, and the Future of Embedded Security
DTSTAMP:20130618T164412Z
DTSTART:20120312T210000Z
DTEND:20120312T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/motors-voters-and-future-embedded-security
LOCATION:CS 1240
DESCRIPTION:Stephen Checkoway PhD Candidate\, University of California\, San Diego    \n Abstract:     The stereotypical view of computing\, and hence computer  \n security\, is a landscape filled with laptops\, desktops\, smartphones and  \n servers\; general purpose computers in the proper sense. However\, this is but  \n the visible tip of the iceberg. In fact\, most computing today is invisibly  \n embedded into systems and environments that few of us would ever think of as  \n computers. Indeed\, applications in virtually all walks of modern life\, from  \n automobiles to medical devices\, power grids to voting machines\, have evolved  \n to rely on the same substrate of general purpose microprocessors and  \n (frequently) network connectivity that underlie our personal computers. Yet  \n along with the power of these capabilities come the same potential risks as  \n well. My research has focused on understanding the scope of such problems by  \n exploring vulnerabilities in the embedded environment\, how they arise\, and  \n the shape of the attack surfaces they expose. In this talk\, I will  \n particularly discuss recent work on two large-scale platforms: modern  \n automobiles and electronic voting machines. In each case\, I will explain how  \n implicit or explicit assumptions in the design of the systems have opened  \n them to attack. I will demonstrate these problems\, concretely and completely\,  \n including arbitrary control over election results and remote tracking and  \n control of an unmodified automobile. I will explain the nature of these  \n problems\, how they have come to arise\, and the challenges in hardening such  \n systems going  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9058.field_date.0.67
SUMMARY:AI In the Wild - Can we measure value in the brain?
DTSTAMP:20130618T164412Z
DTSTART:20120313T180000Z
DTEND:20120313T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/ai-wild-can-we-measure-value-brain
LOCATION:1240cs
DESCRIPTION:AISEM presents Can we measure value in the brain? Adrianna Teriakidis  \n Post-Doctoral Researcher working with Rick Jenison\, Department of Psychology  \n A fundamental question in Neuroeconomics is how and where in the brain value  \n is computed for decision-making. Activity in the amygdala\, although  \n classically associated with emotions\, has been shown recently to reflect  \n value at the time of decision making. We record electrophysiology from the  \n human amygdala of awake\, behaving patients undergoing diagnosis for  \n pharmacologically intractable epilepsy. We use neuron spiking rates recorded  \n from the amygdala to predict a patient-participant's revealed choice on a  \n trial-by-trial basis. We can also modulate these choices using sub-threshold  \n electrical stimulation. Bio: Adrianna Teriakidis received her PhD from the  \n University of Edinburgh for the thesis "Intra-neuronal influences on  \n development of the mammalian neuromuscular junction". This talk is part of  \n the AI In the Wild series. For more information\, including upcoming talks\,  \n click here.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9090.field_date.0.68
SUMMARY:Bilge Mutlu: Professors & Pizza - HCI
DTSTAMP:20130618T164412Z
DTSTART:20120313T210000Z
DTEND:20120313T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/professors-pizza-hci
LOCATION:1240
DESCRIPTION:Not sure which CS class to take next? Come find out about cool research being  \n done in Human-Computer Interaction while eating free pizza! Watch this video  \n about the research our speaker\, Professor Bilge Mutlu\, has been doing with  \n robots.   Abstract: Robots are unique in their ability to afford interaction  \n using the wider range of human communicative capabilities. When used  \n effectively in human interactions\, these capabilities promise significant  \n positive social\, cognitive\, and task outcomes. How might robots take full  \n advantage of this promise to improve our lives? In this talk\, I will describe  \n a research program aimed at building a computational understanding of these  \n human communicative capabilities and using this understanding to design  \n effective social robots.   Speaker Information: Bilge Mutlu is an assistant  \n professor of computer science\, psychology\, and industrial engineering at the  \n University of Wisconsin–Madison. At Wisconsin\, he directs a research  \n program on designing robotic technologies that improve how people learn\,  \n communicate\, and work. Dr. Mutlu is the recipient of three Best Paper Awards  \n (HRI 2008\, 2009\, 2011)\, the NSF CAREER Award\, and Fulbright Fellowship.    \n Presented by The Hub
END:VEVENT
BEGIN:VEVENT
UID:calendar.9091.field_date.0.69
SUMMARY:Sushmita Roy: Computational Approaches to Modeling Gene Regulation in One and  \n Many Species
DTSTAMP:20130618T164412Z
DTSTART:20120313T210000Z
DTEND:20120313T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/computational-approaches-modeling-gene-regulation-one-and-many-species
LOCATION:Biotechnology Center...
DESCRIPTION:  Changes in gene regulation are hypothesized to play a major role in  \n adaptation and evolution of organisms. In recent years\, functional genomics  \n approaches have been used to measure different aspects of the gene regulation  \n machinery in single species and extended to multiple species. These  \n functional genomics datasets give us the unique opportunity to develop more  \n comprehensive models of gene regulation. However\, doing so requires us to  \n develop novel computational tools that integrate such datasets within one  \n species and across multiple species. In this talk\, I will present  \n computational methods to integrate different types of datasets for the  \n regulatory network of the model organism\, Drosophila melanogaster. We show  \n that data integration is key to improved performance and increased coverage  \n of the fly regulatory network.      I will then describe a multi-species  \n analysis framework\, which comprises a (1) novel multi-species clustering  \n algorithm\, Arboretum that identifies modules in species across large  \n evolutionary distances\, and (2) a set of metrics to examine patterns of  \n conservation and divergence in these modules\, as well as the factors that  \n drive divergence. We applied our approach to expression profiles measured in  \n 8 species of Ascomycota fungi under glucose depletion and under heat shock.  \n In both responses\, the transcriptomes are captured by five conserved  \n expression modules\, however\, the degree of gene content conservation in the  \n module was substantially lower in heat shock than glucose depletion\,  \n suggesting a stronger conservation of the latter response.   Our approaches  \n for integrating different types of regulatory datasets\, within one organism  \n and across multiple organisms\, can lead us to systematically understand the  \n structure\, function\, and evolution of regulatory networks.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9029.field_date.0.70
SUMMARY:Robert L. Bocchino Jr.: Deterministic Parallel Java
DTSTAMP:20130618T164412Z
DTSTART:20120315T210000Z
DTEND:20120315T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/deterministic-parallel-java
LOCATION:1240 CS
DESCRIPTION:Speaker: Robert L. Bocchino Jr.\, Carnegie Mellon University Cookies at  \n 3:30pm\; talk begins at 4:00pm. Abstract: With the proliferation of multicore  \n computers on the desktop\, parallel programming is becoming widespread\, but it  \n is still hard. One way to make it easier is to ensure that parallel programs  \n are deterministic: that is\, they produce the same output on every execution  \n with a given input\, regardless of the parallel schedule chosen. Determinism  \n makes parallel programs much easier to write\, understand\, debug\, and  \n maintain. Further\, many (though not all) parallel programs are intended to be  \n deterministic. However\, general-purpose languages — particularly those that  \n allow updates to shared data\, pointers\, and aliasing — typically do not  \n guarantee determinism. Instead\, they require programmers to use testing\,  \n inspection\, or some other method to ensure that their programs behave  \n deterministically. In this talk\, I will present my Ph.D. thesis work on  \n Deterministic Parallel Java (DPJ). DPJ is an explicitly parallel language  \n based on Java. Like Java\, it allows shared mutable objects and aliasing.  \n However\, it uses programmer annotations called effects to guarantee  \n determinism by default: the program is guaranteed to run deterministically  \n unless the programmer explicitly requests a controlled form of  \n nondeterminism. Further\, any nondeterminism is subject to strong safety  \n guarantees\, including freedom from data races and an intuitive way to compose  \n deterministic and nondeterministic code. Finally\, DPJ supports the  \n encapsulation of common parallel patterns into easy-to-use frameworks that  \n provide strong correctness guarantees. I will give an overview of the typing  \n mechanisms that DPJ uses to provide its guarantees\, and I will illustrate  \n some important parallel patterns that DPJ can express with good performance.  \n I will conclude with a brief summary of work currently in progress on (1)  \n making DPJ-style effect checking more flexible and powerful by adding a novel  \n idea called “local uniqueness” and (2) integrating DPJ-style effect  \n checking with Hoare-style reasoning\, particularly for verifying framework  \n implementations\, where extra power is needed beyond what DPJ provides. Bio:  \n Robert Bocchino is a Postdoctoral Associate at Carnegie Mellon University and  \n is supported by a CRA/CCC Computing Innovation Fellowship. His research  \n interests lie in programming language design\, type theory\, formal  \n verification\, and concurrency. Robert completed his Ph.D. at the University  \n of Illinois at Urbana-Champaign in fall 2010. His dissertation on  \n Deterministic Parallel Java received the 2010 Outstanding Dissertation Award  \n from ACM SIGPLAN and the 2012 David J. Kuck Outstanding Thesis Award from the  \n University of Illinois. At CMU\, Robert is working with Jonathan Aldrich on  \n the design and verification of high-level parallel abstractions using a  \n combination of DPJ-style effects\, unique references\, and program logic. He  \n has also contributed to the design of the type system in the Plaid language  \n project. Plaid makes object states and transitions a first-class feature of  \n the language\, and its type system uses unique and immutable references to  \n reason about object aliases.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9055.field_date.0.71
SUMMARY:Foundations of Dynamic Access Control
DTSTAMP:20130618T164412Z
DTSTART:20120316T150000Z
DTEND:20120316T160000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/foundations-dynamic-access-control
LOCATION:2310
DESCRIPTION:Speaker: Prasad Naldurg (Microsoft Research\, India) Title: Foundations of  \n Dynamic Access Control Abstract: In this talk\, I will describe our work on  \n understanding the foundations of dynamic access control. In contrast to  \n traditional operating systems\, new commercial operating systems e.g.\, Windows  \n 7\, and research operating systems such as Asbestos and Flume\, include labels  \n for integrity protection. Unlike the strict Bell-LaPadula mandatory access  \n controls\, these labels are allowed to change in controlled ways by users and  \n applications. The implications of these dynamic changes need to be examined  \n carefully\, and existing formalisms cannot express or help us understand their  \n impact on access control safety. We present a logic-programming framework to  \n specify\, analyze and automatically verify such dynamic access control models.  \n We study the problem of reachability (equivalently safety) in these models  \n and show that they are undecidable in the general case. We also identify a  \n reasonably expressive fragment of this formalism that has a sound and  \n complete decision procedure. We build a theory (and tool) based on bounded  \n model-checking for reasoning about information flow in the general context\,  \n and show its application on real-world use-cases. We are able to highlight  \n several important vulnerabilities in these models\, as well as suggest design  \n changes that can be provably validated. I will conclude a small discussion on  \n open problems in this framework and future work. This talk summarizes some of  \n our work from FMSE 2006\, CCS 2008\, PLAS 2009 (best paper) and SACMAT 2011.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9113.field_date.0.72
SUMMARY:Erez Lieberman Aiden: Zooming Out: New Tools for Probing the Historical  \n Record and the Human Genome
DTSTAMP:20130618T164412Z
DTSTART:20120319T170000Z
DTEND:20120319T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/zooming-out-new-tools-probing-historical-record-and-human-genome
LOCATION:Union South Northwoods (...
DESCRIPTION:  Abstract: New structures often emerge when we explore a known phenomenon  \n from a more global vantage point. For instance\, any given book can be read  \n and comprehended. But what happens when we try to read all the books at once?  \n Or: the local structure of DNA is a double helix. But if DNA did not fold  \n further\, the human genome - which is two meters long - could never fit inside  \n the nucleus of a cell. How does it fold? This talk will focus on the  \n extraordinary potential of technologies that enable us to zoom out\, in the  \n process transforming familiar concepts\, like the contents of a book or the  \n shape of DNA\, into new research horizons. First\, I will describe efforts\,  \n together with my collaborator Jean-Baptiste Michel and Google\, to create  \n tools for the quantitative analysis of a significant portion of the  \n historical record. We began by constructing a reliable corpus of digitized  \n texts containing about 4% of all books ever printed. Analysis of this corpus  \n enables us to investigate cultural trends quantitatively. 'Culturomics'  \n provides insights about fields as diverse as lexicography\, the evolution of  \n grammar\, collective memory\, the adoption of technology\, the pursuit of fame\,  \n censorship\, and historical epidemiology. The Google Ngram Viewer\, a simple  \n web-based tool we released for the analysis of this corpus\, was used over a  \n million times in the first 24 hours. Culturomics extends the boundaries of  \n rigorous quantitative inquiry to a wide array of new phenomena spanning the  \n social sciences and the humanities. In the second half of my talk\, I will  \n describe Hi-C\, a novel technology for probing the three-dimensional  \n architecture of whole genomes. Developed together with collaborators at the  \n Broad Institute and UMass Medical School\, Hi-C couples proximity-dependent  \n DNA ligation and massively parallel sequencing. My lab employs Hi-C to  \n construct spatial proximity maps of the human genome. Hi-C maps have revealed  \n that active and inactive portions of the human genome are spatially  \n segregated\, ie\, that cells employ a sort of 'regulatory origami' as they turn  \n genes on and off. At the megabase scale\, the genomic fold is consistent with  \n a fractal globule\, a knot-free conformation that enables maximally dense  \n packing while preserving the ability to easily fold and unfold any genomic  \n locus. Brief bio: Erez Lieberman Aiden is a fellow at the Harvard Society of  \n Fellows and Visiting Faculty at Google. His work integrates mathematical and  \n physical theory with the invention of new technologies. He recently invented  \n a method for three-dimensional genome sequencing\; he subsequently led the  \n team that\, in 2009\, reported the first three dimensional map of the human  \n genome. Together with collaborator Jean-Baptiste Michel\, he developed  \n culturomics\, a quantitative approach to the study of history and culture that  \n relies on computational analysis of a significant fraction of the historical  \n record. This work led to the creation of the Google Ngram Viewer\, a  \n supplemental website that was visited over a million times in the 24 hours  \n after its launch. Erez's research has won numerous awards\, including an NIH  \n New Innovator Award\; the GE & Science Prize for Young Life Scientists\; the  \n Lemelson-MIT prize for the best student inventor at MIT\; the American  \n Physical Society's Award for the Best Doctoral Dissertation in Biological  \n Physics\; recognition for one of the top 20 "Biotech Breakthroughs that will  \n Change Medicine"\, by Popular Mechanics\; and membership in Technology Review's  \n 2009 TR35 recognizing the top 35 innovators under 35.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9082.field_date.0.73
SUMMARY:Ranjita Bhagwan: Automating Network and Systems Diagnostics
DTSTAMP:20130618T164412Z
DTSTART:20120319T183000Z
DTEND:20120319T193000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/automating-network-and-systems-diagnostics
LOCATION:CS 2310
DESCRIPTION:Network and systems diagnostics are complex problems that are currently  \n handled mostly through manually constructed procedures and knowledgebases. In  \n spite of these problems being pervasive\, current techniques are unable to  \n resolve a significant fraction of them in acceptable turnaround time. Our  \n objectives through the NetPrints\, Déjà vu\, and Baaz projects has been to  \n hone in onto some aspects of these problems\, and automatically find solutions  \n through the use of various machine learning techniques in a timely manner.  \n NetPrints and Déjà vu concentrate on building knowledgebases for diagnosing  \n networking-related problems\, while Baaz finds misconfigurations in systems  \n access control\, thereby mitigating the problem of insider attacks. In this  \n talk\, I will describe the objectives\, design and evaluation of these systems  \n and how they aid network and system administrators effectively find the  \n root-cause of problems and fix them.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9018.field_date.0.74
SUMMARY:Paul Valiant: The Challenge of Data Efficiency: Vignettes from Statistics and  \n Evolution
DTSTAMP:20130618T164412Z
DTSTART:20120319T210000Z
DTEND:20120319T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/challenge-data-efficiency-vignettes-statistics-and-evolution
LOCATION:1240
DESCRIPTION:The Challenge of Data Efficiency: Vignettes from Statistics and Evolution As  \n the amount of data that can be brought to bear on a task grows\, so too does  \n the difficulty and importance of making the best use of our data. In this  \n talk we will consider two settings of this problem\, one from statistics and  \n one from the natural world. A basic question in statistics is to determine  \n how much data is needed to estimate various properties of one or more  \n probability distributions. We consider the fundamental properties of support  \n size\, entropy\, and the distance between two distributions. We present a  \n unified new approach to these problems which yields the first explicit  \n sublinear algorithms: given one or more probability distributions on a domain  \n of size n\, we show how to estimate each of these properties using n/log n  \n samples\, where n corresponds to the support size of the distribution. Our  \n approach leverages the techniques of linear programming to yield a single  \n algorithm to estimate all three properties\, and\, indeed\, all properties from  \n a wider class that contains them. Further\, we construct a new discrete  \n version of the multivariate central limit theorem to show that n/log n  \n samples are in fact needed to estimate these properties and thus show that  \n our algorithm uses the optimal number of samples\, to within constant factor.  \n In the second half of the talk\, we take our inspiration from biological  \n evolution -- a natural process that\, from vast amounts of data and complexity  \n yields surprisingly effective results. In this talk we formulate a notion of  \n "evolvability" for functions with domain and range that are real-valued  \n vectors\, an appropriate way of expressing many natural biological processes  \n such as protein expression regulation. We show that linear and fixed-degree  \n polynomial functions are evolvable in the following dually robust sense:  \n There is a single evolution algorithm that for all convex loss functions  \n converges for all distributions. We further examine the scope of the  \n evolvability model and discuss future directions towards a computational  \n understanding of evolution. Bio: Paul Valiant attended Stanford from  \n 2000-2004\, earning a BS in Mathematics\, a BS in Physics\, and an MS in  \n Computer Science. From 2004 – 2008 he studied at MIT\, earning a PhD in  \n Computer Science. After that he has been at UC Berkeley on an NSF  \n Mathematical Sciences Postdoctoral Research Fellowship. His awards include  \n the Machtey Award (Best Student Paper) at FOCS 2005 (co-winner)\, Best Student  \n Paper Award at the Theory of Cryptography Conference 2008\, Stanford  \n Mathematics Department Research Award for Undergraduate Honors Thesis on  \n “General Relativity”. He was a Gold Medalist at the International  \n Mathematical Olympiad in 1999. A three-time member of the US International  \n Mathematical Olympiad team\, he taught last summer at the US Math Olympiad  \n Summer Program training the US team. His research interests span the areas of  \n Statistics and the quest for data efficiency\, Security and game theory\, Fluid  \n dynaimics-the Navier-Stokes problem\, and Protein folding and evolution.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9089.field_date.0.75
SUMMARY:Alexander Fish: Channel Estimation in Wireless Communication in Almost Linear  \n Time.
DTSTAMP:20130618T164412Z
DTSTART:20120319T210000Z
DTEND:20120319T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/channel-estimation-wireless-communication-almost-linear-time
LOCATION:Wisconsin Institute for...
DESCRIPTION:We will present the model of mobile communication\, and will discuss the  \n problem of channel estimation -- finding time-frequency shifts which a  \n waveform undergoes while transmitted in the presence of a white noise. The  \n digital model of the problem involves signals of length N (complex-valued  \n vectors of length N). The current method of solving digital channel  \n estimation problem uses O(N^2 log(N)) arithmetic operations. Using ideas from  \n representation theory\, we will present a new method of solving channel  \n estimation problem of complexity O(N log(N)). The applications of the new  \n method to mobile communication and GPS system will be  \n discussed.<?xml:namespace prefix = o />   This is a joint work with  \n S.Gurevich (Math\, UW-Madison)\, R.Hadani (Math\, UT-Austin)\, A.Sayeed (ECE\,  \n UW-Madison)\, O.Schwartz (EECS\, UC Berkeley).  Coffee and tea will be served  \n in conjunction with WID Discovery at 3:30pm
END:VEVENT
BEGIN:VEVENT
UID:calendar.9092.field_date.0.76
SUMMARY:Joel Hestness: Netrace
DTSTAMP:20130618T164412Z
DTSTART:20120320T210000Z
DTEND:20120320T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/netrace
LOCATION:CS 1221
DESCRIPTION:Given that modern and future compute devices will integrate a growing number  \n of components on a single chip\, the importance of high-performance and  \n low-power on-chip communication is on the rise. However\, given the diversity  \n and complexity of components on these highly-integrated chips\, investigating  \n chip-level architecture designs aimed at achieving these performance and  \n power objectives is very difficult. Current and future systems will see  \n increasing and diversifying communication demands as the systems and common  \n applications change to meet the demands of consumers. In this talk\, I will  \n discuss the current state-of-the-art in on-chip communication performance and  \n power evaluation methodologies\, including my prior work\, Netrace\, which is a  \n trace-based\, dependency-driven approach to evaluating interconnect fabrics of  \n new and emerging chips. In addition to motivating these techniques by  \n pointing out their benefits\, I will describe their disadvantages\, and  \n opportunities to improve future evaluation techniques. For more information  \n about Netrace: http://www.cs.utexas.edu/~netrace/ Joel Hestness is a PhD  \n student transferring to UW from the University of Texas at Austin\, where he  \n started is PhD working with Prof. Steve Keckler. Joel's research focuses on  \n on-chip communication and networks\, and more recently on heterogeneous  \n computing chips. For more information\, check out his site:  \n http://www.cs.utexas.edu/~hestness/
END:VEVENT
BEGIN:VEVENT
UID:calendar.9094.field_date.0.77
SUMMARY:Lisa Grueneberg: WACM Speaker Series: Lisa Grueneberg from Epic
DTSTAMP:20130618T164412Z
DTSTART:20120320T220000Z
DTEND:20120320T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/wacm-speaker-series-lisa-grueneberg-epic
LOCATION:2310
DESCRIPTION:Lisa Grueneberg is a graduate of the UW CS department from December of 1993.  \n Following graduation\, she happened upon an ad in the WI State Journal for a  \n programmer position at a small local healthcare software company. Five  \n thousand employees and 17 years later\, she continues to work at Epic doing  \n her part to improve the way healthcare is delivered. She has held a variety  \n of technical positions in the company\, and is currently working on  \n internationalization and localization efforts as Epic's customer base expands  \n outside of the United States.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9124.field_date.0.78
SUMMARY:Russ Greiner: AI Seminar: WebIC: An Effective "Complete-Web" Recommender  \n System
DTSTAMP:20130618T164412Z
DTSTART:20120322T171500Z
DTEND:20120322T181500Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/ai-seminar-webic-effective-complete-web-recommender-system
LOCATION:CS 1240
DESCRIPTION:WebIC: An Effective "Complete-Web" Recommender System Many web recommendation  \n systems direct users to webpages\, from a single website\, that other similar  \n users have visited. By contrast\, our WebIC web recommendation system is  \n designed to locate "information content (IC) pages" --- pages the current  \n user needs to see to complete her task --- from essentially anywhere on the  \n web. WebIC first extracts the "browsing properties" of each word encountered  \n in the user's current click-stream --- eg\, how often each word appears in the  \n title of a page in this sequence\, or in the "anchor" of a link that was  \n followed\, etc. It then uses a user- and site-independent model\, learned from  \n a set of annotated web logs acquired in a user study\, to determine which of  \n these words is likely to appear in an IC page. We discuss how to use these  \n IC-words to find IC-pages\, and demonstrate empirically that this  \n browsing-based approach works effectively. Bio: After earning a PhD from  \n Stanford\, Russ Greiner worked in both academic and industrial research before  \n settling at the University of Alberta\, where he is now a Professor in  \n Computing Science and the founding Scientific Director of the Alberta  \n Ingenuity Centre for Machine Learning\, which won the ASTech Award for  \n "Outstanding Leadership in Technology" in 2006. He has been Program Chair for  \n the 2004 "Int'l Conf. on Machine Learning"\, Conference Chair for 2006 "Int'l  \n Conf. on Machine Learning"\, Editor-in-Chief for "Computational Intelligence"\,  \n and is serving on the editorial boards of a number of other journals. He was  \n elected a Fellow of the AAAI (Association for the Advancement of Artificial  \n Intelligence) in 2007\, and was awarded a McCalla Professorship in 2005-06 and  \n a Killam Professorship in 2007.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9004.field_date.0.79
SUMMARY:TBA
DTSTAMP:20130618T164412Z
DTSTART:20120322T203000Z
DTEND:20120322T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/tba
LOCATION:1240
END:VEVENT
BEGIN:VEVENT
UID:calendar.9088.field_date.0.80
SUMMARY:Alekh Agarwal (WID Optimization Faculty Candidate): Statistics meets  \n Computation: Exploiting structure for efficient learning with large data-sets
DTSTAMP:20130618T164412Z
DTSTART:20120322T210000Z
DTEND:20120322T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/statistics-meets-computation-exploiting-structure-efficient-learning-large-data-sets
LOCATION:CS room 1240
DESCRIPTION:The past decade has seen the emergence of datasets of an unprecedented scale\,  \n with both large sample sizes and dimensionality. Massive data sets arise in  \n various domains\, among them computer vision\, natural language processing\,  \n computational biology\, social networks analysis and recommendation systems\,  \n to name a few.  In many such problems\, the bottleneck is not just the number  \n of data samples\, but also the computational resources available to process  \n the data.  An important goal in such problems is to understand how we might  \n leverage the statistical structure of the problem to get efficient  \n computational procedures.<?xml:namespace prefix = o />   In this talk\, I  \n present three research threads that provide different lines of attack on this  \n broader research agenda: (i) distributed algorithms for statistical  \n inference\; (ii) interplay between statistical and computational complexities  \n in structured high-dimensional estimation\; and (iii) some computational  \n trade-offs in noisy matrix decomposition. In the first two examples\, we will  \n demonstrate how we can beat the worst-case complexity of the computational  \n problem by exploiting the statistical structure. The last part will show an  \n example where we get more and more computationally expensive solutions\, based  \n on the statistical demands we place on the algorithm.   [Joint work with  \n John Duchi\, Sahand Negahban\, Peter Bartlett and Martin Wainwright
END:VEVENT
BEGIN:VEVENT
UID:calendar.9119.field_date.0.81
SUMMARY:Russ Greiner: Budgeted Learning of Effective Classifiers
DTSTAMP:20130618T164412Z
DTSTART:20120323T170000Z
DTEND:20120323T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/budgeted-learning-effective-classifiers
LOCATION:140 Bardeen\, 1215 Linden...
DESCRIPTION:Abstract: Researchers often use clinical trials to collect the data needed to  \n evaluate some hypothesis\, or produce a classifier. During this process\, they  \n have to pay the cost of performing each test. Many studies will run a  \n comprehensive battery of tests on each subject\, for as many subjects as their  \n budget will allow - ie\, "round robin" (RR). We consider a more general model\,  \n where the researcher can sequentially decide which single test to perform on  \n which specific individual\; again subject to spending only the available  \n funds. Our goal here is to use these funds most effectively\, to collect the  \n data that allows us to learn the most accurate classifier. We first explore  \n the simplified "coins version" of this task. After observing that this is  \n NP-hard\, we consider a range of heuristic algorithms\, both standard and  \n novel\, and observe that our "biased robin" approach is both efficient and  \n much more effective than most other approaches\, including the standard RR  \n approach. We then apply these ideas to learning a naive-bayes classifier\, and  \n see similar behavior. Finally\, we consider the most realistic model\, where  \n both the researcher gathering data to build the classifier\, and the user (eg\,  \n physician) applying this classifier to an instance (patient) must pay for the  \n features used - eg\, the researcher has $10\,000 to acquire the feature values  \n needed to produce an optimal $30/patient classifier. Again\, we see that our  \n novel approaches are almost always much more effective that the standard RR  \n model. (Joint work with Aloak Kapoor\, Dan Lizotte and Omid Madani.)
END:VEVENT
BEGIN:VEVENT
UID:calendar.9103.field_date.0.82
SUMMARY:Daniel Dadush (Optimization Faculty Candidate): Recent Progress on Algorithms  \n and Cutting Planes for Convex Integer Programs
DTSTAMP:20130618T164412Z
DTSTART:20120326T210000Z
DTEND:20120326T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/recent-progress-algorithms-and-cutting-planes-convex-integer-programs
LOCATION:CS room 1240
DESCRIPTION:Integer Programming (IP) is the one of the most important and effective  \n modeling paradigms for optimization problems in operations research. Though  \n modern IP solvers can now efficiently solve a large class of IP models\, many  \n salient models\, in particular those including non-linear constraints and  \n objectives remain intractable.   <?xml:namespace prefix = o />   In this  \n talk\, I will survey recent theoretical progress for the class of convex  \n integer programs\, i.e. the class of IPs whose continuous relaxation is a  \n general convex program. Convex IPs naturally occur as convexifications of  \n general non-linear programs\, and hence are important to study from the  \n perspective of obtaining strong bounds for non-linear problems. Cutting  \n planes\, i.e. valid linear inequalities for integer points inside the feasible  \n region\, are one of the most useful tools for solving real world IPs. The  \n properties of cutting planes for non-linear IPs however remain poorly  \n understood. In the first part of the talk\, I will describe recent results on  \n cutting plane closures for Convex IPs. Firstly\, we show that the  \n Chvatal-Gomory closure of any compact convex set is rational polyhedral. This  \n resolves a long standing question of Schrijver for irrational polytopes.  \n Second\, we show that the split closure of any strictly convex body is  \n finitely generately\, though need not be polyhedral.   In the second part of  \n the talk\, I discuss improvements to the computational complexity of the  \n Integer Programming Problem. In the seminal works of Lenstra (MOR `83) and  \n Kannan (MOR `87)\, it was shown that any n variable Integer Linear Program  \n (ILP) can be solved in time O(n)^{2.5n} (with polynomial dependence on the  \n remaining parameters). In recent work\, I give an O(n)^n time algorithm to  \n minimize a convex function over the integer points in any n dimensional  \n convex body. This yields the first exact algorithm for Convex IP and the  \n current fastest algorithm for ILP. The algorithm relies on new insights in  \n the geometry of numbers as well as new techniques for lattice problems.   \n Coffee will be served at 3:30pm<?xml:namespace prefix = v /><?xml:namespace  \n prefix = w />  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9093.field_date.0.83
SUMMARY:Emily Blem: ISA Wars
DTSTAMP:20130618T164412Z
DTSTART:20120327T210000Z
DTEND:20120327T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/isa-wars
LOCATION:CS 1221
DESCRIPTION:RISC vs. CISC wars raged in the 80s and its debatable if the question was  \n ever settled. With the proliferation of tablets\, smartphones which run ARM (a  \n RISC ISA) compared to desktops\,laptops which run x86 (a CISC ISA)\, the  \n question of whether ISA plays a role in performance or energy efficiency is  \n becoming important again. In this talk\, we examine and analyse in detail  \n several implementations of the ARM and x86 ISA to provide a definitive answer  \n to this decades old question of RISC vs. CISC. Much of this work is on-going  \n and feedback (aka criticism) on all aspects of this work are welcome
END:VEVENT
BEGIN:VEVENT
UID:calendar.9134.field_date.0.84
SUMMARY:Kendrick Boyd: Precision-Recall Space and Empirical Algorithm Evaluation
DTSTAMP:20130618T164412Z
DTSTART:20120327T210000Z
DTEND:20120327T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/precision-recall-space-and-empirical-algorithm-evaluation
LOCATION:Biotechnology Center...
DESCRIPTION:  Abstract: ROC curves are widely used to represent the quality of medical  \n screening procedures such as mammography. We show that in screening for  \n diseases with rare prevalence\, such as mammography for breast cancer\,  \n precision-recall (PR) curves have some significant advantages over ROC  \n curves. Because of these advantages\, PR curves\, and the areas under them\, are  \n already the evaluation metrics of choice for other tasks characterized by low  \n prevalence\, such as information retrieval. While PR curves are frequently  \n used as a simple replacement for ROC curves\, there are subtleties regarding  \n PR curves that must be considered. It is already known that PR curves vary as  \n class skew varies. What was not recognized before is that there is a region  \n of PR space that is completely unachievable\, and the size of this region  \n varies only with the skew. We precisely characterize the size of the  \n unachievable region and discuss its implications for empirical evaluation  \n methodology in machine learning.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9131.field_date.0.85
SUMMARY:Elizabeth Bodine-Baron (Optimization Faculty Candidate): From Distributed  \n Search to Matching Markets: the Advantage of Social Network Structure
DTSTAMP:20130618T164412Z
DTSTART:20120329T193000Z
DTEND:20120329T203000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/distributed-search-matching-markets-advantage-social-network-structure
LOCATION:Engineering Hall Room...
DESCRIPTION:In this talk\, we examine two distinct problems\, distributed search and  \n matching markets with externalities\, through the lens of social networks.  \n  In the first\, we focus on the problem of distributed search in social  \n networks - forwarding a message using only local contact information.  In  \n order to model networks that are "searchable" by such an algorithm\, we  \n construct a generalization of stochastic Kronecker graphs for generating  \n random social networks\, introducing a Kronecker-like operator and defining a  \n family of generator matrices dependent on distances between nodes in a  \n specific graph embedding.  Using this model\, we highlight a few of our  \n results on the performance of greedy message forwarding algorithms and  \n demonstrate that "distance-dependent Kronecker graphs" can generate  \n searchable networks. <?xml:namespace prefix = o /> In the second problem\, we  \n focus on matching markets\, such as those used to match medical interns to  \n hospital residencies and assign housing to college students.  Externalities  \n such as complementarities and peer effects can severely complicate the  \n preference ordering of each agent and lead to serious problems for market  \n stability.  We note that peer effects are often the result of underlying  \n social connections\, and so we explore a formulation of the market where peer  \n effects are derived from an underlying social network. The key feature of our  \n model is that it captures peer effects and complementarities using utility  \n functions\, rather than traditional preference ordering. With this model and  \n considering a slightly different notion of stability\, we prove that stable  \n matchings always exist and characterize the set of stable matchings in terms  \n of social welfare.  To characterize the efficiency of matching markets with  \n externalities\, we provide "price of stability" and "price of anarchy" bounds  \n and find that the structure of the social network (e.g. how well clustered  \n the network is) plays a large role.  Finally\, we demonstrate the performance  \n of two matching algorithms on a real-world matching problem: assigning  \n students to housing at Caltech while incorporating their social network  \n structure.      
END:VEVENT
BEGIN:VEVENT
UID:calendar.9132.field_date.0.86
SUMMARY:Aarti Singh: Leveraging Information Structure to Overcome Data Deficiencies  \n in Large Complex Systems
DTSTAMP:20130618T164412Z
DTSTART:20120329T210000Z
DTEND:20120329T210000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/leveraging-information-structure-overcome-data-deficiencies-large-complex-systems
LOCATION:Room 1240
DESCRIPTION:  Despite great advances in high-throughput technologies\, the scale and  \n complexity of modern systems makes it impossible to perfectly monitor them.  \n As a result\, modern datasets are severely under-sampled\, corrupted by noise  \n and outliers\, high-dimensional\, and unordered. However\, the information of  \n interest is often structured (in the form of clusters\, bi-clusters\, graphs  \n and other topological properties). Leveraging this information structure is  \n key to overcoming the data deficiencies\, and enabling robust and  \n resource-efficient inference in large complex systems. In this talk\, I will  \n demonstrate how information structure in the form of clusters can be  \n extracted from large-scale\, noisy and incomplete data. First\, I will  \n characterize robustness of a popular spectral clustering algorithm\, and  \n establish its near-optimality using information theoretic lower bounds. For  \n large-scale datasets\, it might be prohibitive to obtain or compute all the  \n similarities. To address this\, I will present a novel framework for "active"  \n hierarchical clustering that uses few\, selectively sampled similarities.  \n Coupled with the robustness analysis\, this yields a new efficient  \n hierarchical spectral clustering method that only requires O(N log^2 N)  \n selective similarities\, instead of all O(N^2) pairwise similarities\, to  \n cluster N objects and runs in linear time. Finally\, I will briefly mention  \n robustness results for high dimensional clustering settings\, where the  \n clusters are small and characterized by only a few relevant features. Bio:  \n Aarti Singh is an Assistant Professor in the Machine Learning Department at  \n Carnegie Mellon University. She received a Ph.D. degree in Electrical  \n Engineering from the University of Wisconsin-Madison in 2008 and was a  \n Postdoctoral Research Associate at the Program in Applied and Computational  \n Mathematics at Princeton University from 2008-2009. Her research brings  \n together tools from machine learning\, statistics and signal processing to  \n develop theoretically sound and practically feasible methods for inference in  \n large complex systems\; with applications to sensor networks\, epidemiology\,  \n drug-protein interaction discovery\, and brain networks.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9140.field_date.0.87
SUMMARY:Moshe Y. Vardi: From Philosophical to Industrial Logics
DTSTAMP:20130618T164412Z
DTSTART:20120403T160000Z
DTEND:20120403T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/philosophical-industrial-logics
LOCATION:B130 Van Vleck
DESCRIPTION:One of the surprising developments in the area of program verification is how  \n several ideas introduced by logicians in the first part of the 20th century  \n ended up yielding at the start of the 21st century industry-standard  \n property-specification languages called PSL and SVA. This development was  \n enabled by the equally unlikely transformation of the mathematical machinery  \n of automata on infinite words\, introduced in the early 1960s for second-order  \n arithmetics\, into effective algorithms for industrial model-checking tools.  \n This talk attempts to trace the tangled threads of this development. This  \n talk is part of the North American Annual Meeting of the Association for  \n Symbolic Logic (http://www.math.wisc.edu/~asl2012/slides.htm).
END:VEVENT
BEGIN:VEVENT
UID:calendar.9059.field_date.0.88
SUMMARY:Kayur Patel: Faculty Candidate Talk: Kayur Patel
DTSTAMP:20130618T164412Z
DTSTART:20120409T203000Z
DTEND:20120409T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/faculty-candidate-talk-kayur-patel
LOCATION:1240 CS
DESCRIPTION:Lowering the Barrier to Applying Machine Learning Kayur Patel\, University of  \n Washington Abstract: Data is driving the future of computation: analysis\,  \n visualization and learning algorithms power systems that help us diagnose  \n cancer\, live sustainably\, and understand the universe. Yet\, the data  \n explosion has outstripped our tools to process it\, leaving a gap between  \n powerful new algorithms and what real programmers can apply in practice. I  \n examine how data affects the way we program. My current research focuses on  \n machine learning algorithms. I found that the key barrier to adoption is not  \n a poor understanding of the machine learning algorithms themselves\, but  \n rather a poor understanding of the process for applying those algorithms and  \n poor tool support for that process. I have created new programming and  \n analysis tools that support programmers by helping them (1) implement machine  \n learning systems and analyze results\, (2) debug data and (3) design and track  \n experiments. Bio: Kayur Patel is a Ph.D. student in the Department of  \n Computer Science and Engineering at the University of Washington. Kayur  \n received an M.S in Computer Science from Stanford and a B.S. in Computer  \n Science and Human-Computer Interaction at Carnegie Mellon University. His  \n work has been funded by grants from the NSF and google as well as the NDSEG  \n and Microsoft Research fellowships. Kayur’s research interests are in  \n human-computer interaction\, software engineering\, machine learning and  \n information visualization. For his thesis\, he is studying how programmers  \n apply machine learning to solve problems and build software. Guided by this  \n research\, he creates new development tools that help programmers more  \n effectively use machine learning algorithms.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9087.field_date.0.89
SUMMARY:Ilya Safro: Multiscale Methods for Discrete Optimization Problems
DTSTAMP:20130618T164412Z
DTSTART:20120409T210000Z
DTEND:20120409T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/multiscale-methods-discrete-optimization-problems
LOCATION:Wisconsin Institute for...
DESCRIPTION:In many real-world problems\, a big scale gap can be observed between micro-  \n and macroscopic scales of the problem because of the difference in  \n mathematical (engineering\, social\, biological\, physical\, etc.) models and/or  \n laws at different scales. The main objective of the multiscale algorithms is  \n to create a hierarchy of problems\, each representing the original problem at  \n different coarse scales with fewer degrees of freedom. We will talk about  \n different strategies of creating these hierarchies for large-scale discrete  \n optimization problems: linear ordering\, network compression\, graph  \n partitioning\, clustering\, network generation\, and constrained 2D-layout  \n problem. These strategies are inspired by the classical multigrid frameworks:  \n Geometric Multigrid\, Algebraic Multigrid and Full Approximation Scheme. We  \n will present in details a framework for designing linear time Algebraic  \n Multigrid based multiscale algorithm for the linear ordering\, partitioning  \n and clustering problems. Our multiscale methods have proved themselves to be  \n robust both as a first approximation and as more aggressive optimization  \n solvers. The computational results will be presented.Coffee and tea will be  \n served in conjunction with WID Discovery at 3:30pm
END:VEVENT
BEGIN:VEVENT
UID:calendar.9150.field_date.0.90
SUMMARY:Deborah Muganda-Rippchen: Computing Clustered Alignments of Gene-Expression  \n Time Series
DTSTAMP:20130618T164412Z
DTSTART:20120410T210000Z
DTEND:20120410T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/computing-clustered-alignments-gene-expression-time-series
LOCATION:Biotechnology Center...
DESCRIPTION:Abstract: Identifying similarities and differences in expression patterns  \n across multiple time series can provide a better understanding of the  \n relationships among various chemical treatments or the effects induced by a  \n gene knockout. We consider the task of identifying sets of genes that have a  \n high degree of similarity both in their (i) expression profiles within each  \n treatment\, and (ii) changes in expression responses across treatments.  \n Previously\, we developed an approach for aligning time series that computes  \n clustered alignments. In this approach\, an alignment represents the  \n correspondences between two gene expression time series. Portions of one of  \n the time series may be compressed or stretched to maximize the similarities  \n between the two series. A clustered alignment groups genes such that the  \n genes within a cluster share a common alignment\, but each cluster is aligned  \n independently of the others. Unlike standard gene-expression clustering\,  \n which groups genes according to the similarity of their expression profiles\,  \n the clustered-alignment approach clusters together genes that have similar  \n changes in expression responses across treatments. We have now extended the  \n clustered alignment approach to produce multi-level clusterings that identify  \n subsets of genes that have a high degree of similarity both in their (i)  \n expression profiles within each treatment\, and (ii) changes in expression  \n responses across treatments. We examine the validity of this multi-level  \n clustering method by performing a GO-term enrichment analysis of the  \n clusters. Additionally\, we use permutation testing to determine if our  \n clusters that have alignment scores that are unlikely to occur by chance.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9142.field_date.0.91
SUMMARY:Perry Kivolowitz: Professors & Pizza - Debugging
DTSTAMP:20130618T164412Z
DTSTART:20120410T220000Z
DTEND:20120410T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/professors-pizza-debugging
LOCATION:CS1240
DESCRIPTION:Debugging Art\, Science: Neither\, Either or Both? A talk by Perry Kivolowitz  \n When: 5pm on Tuesday\, April 10th Where: CS1240 What: A talk about debugging  \n with FREE IAN'S PIZZA Abstract: Some say debugging is an art and can't be  \n taught. This has been the fallback explanation for why debugging isn't taught  \n since the dawn of programming. In this talk I will show that the art of  \n debugging is based on science. Science can be taught. Using the concepts  \n presented in this talk should greatly accelerate the maturation of your art.  \n About the speaker: Perry Kivolowitz is an odd duck. With nearly 40 years of  \n programming experience in a many domains\, Perry began lecturing on debugging  \n within the Usenix community and Bell Labs in the 1980s. Much of his career  \n has been in the field of digital visual effects from pioneering work in the  \n brand new multimedia field to co-authoring industry dominating tools in the  \n motion picture industry for 20 years. While C and C++ centric\, this talk is  \n applicable to all languages and all levels of programmer.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9064.field_date.0.92
SUMMARY:Jonathan Lipson: AISEM: When Data Don't Speak: Empirical Legal Research and  \n the Example of Bankruptcy Examiners
DTSTAMP:20130618T164412Z
DTSTART:20120411T200000Z
DTEND:20120411T210000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/aisem-when-data-dont-speak-empirical-legal-research-and-example-bankruptcy-examiners
LOCATION:1240cs
DESCRIPTION:  AISEM Presents When Data Don't Speak:  Empirical Legal Research and the  \n Example of Bankruptcy Examiners Jonathan Lipson Foley & Lardner Professor of  \n Law With the advent of modern computer and information technologies\, legal  \n scholarship has undergone a deep transformation. Today\, a large and growing  \n number of legal academics conduct "empirical" scholarship driven by data  \n about individual cases gathered and analyzed using conventional social  \n science techniques. While this form of analysis can provide powerful insight  \n into what happens "on the ground"\, legal academics are hamstrung because  \n their data are often from sources not intended to be used for data analysis  \n (e.g.\, dockets of pleadings in cases)\; may involve multiple data sets with  \n (many) different variables\; may have missing variables\; will often be text  \n (rather than numerically) based\; will often involve skewness or low  \n statistical power\; and so on. In short\, legal academics attempt to use data  \n not meant to be analyzed in these ways\, and which therefore do not "speak" as  \n clearly to the problems we study as we would hope. The goal of this talk is  \n to explain challenges encountered in a particular data-driven project (the  \n pattern in the use of "examiners" in large corporate bankruptcies) and the  \n limits of conventional regression analysis in this context\, with a view  \n toward describing more generally the kinds of machine-learning and artificial  \n intelligence tools that could aid legal scholarship. Click here to view a  \n recent paper by the speaker on the topic: Understanding Failure: Examiners  \n and the Bankruptcy Reorganization of Large Public Companies Click here for a  \n link to Prof. Lipson's biography. Light refreshments will be served at  \n 2:45pm.   This talk is part of the AI In the Wild series.  For more  \n information\, including upcoming talks\, click here.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9144.field_date.0.93
SUMMARY:Ankit Singla: Jellyfish: Networking Data Centers Randomly
DTSTAMP:20130618T164412Z
DTSTART:20120412T183000Z
DTEND:20120412T193000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/jelly%EF%AC%81sh-networking-data-centers-randomly
LOCATION:CS 2310
DESCRIPTION:Industry experience indicates that the ability to incrementally expand data  \n centers is essential. However\, existing high-bandwidth network designs have  \n rigid structure that interferes with incremental expansion. In this talk\, I  \n will discuss our approach of using random graph based networks to overcome  \n this problem. Interestingly\, in addition to easing incremental  \n expansion\, such networks are *more* cost-efficient than a  \n fat-tree\, supporting as many as 25% more servers at full capacity using the  \n same equipment at the scale of a few thousand nodes. This  \n efficiency improvement grows with scale. For once\, we systems folk can have  \n the best of everything. It does require some effort: An unstructured  \n topology brings new challenges in routing\, physical layout\, and wiring\, some  \n of which I will address in this talk. This is joint work with Chi-Yao Hong  \n and Brighten Godfrey (UIUC)\, and Lucian Popa (HP Labs).
END:VEVENT
BEGIN:VEVENT
UID:calendar.9054.field_date.0.94
SUMMARY:Rethinking the Architecture of Warehouse-Scale Computers: Improving  \n Efficiency and Utilization\; Jason Mars\, University of Virginia
DTSTAMP:20130618T164412Z
DTSTART:20120412T210000Z
DTEND:20120412T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/rethinking-architecture-warehouse-scale-computers-improving-efficiency-and-utilization-jason-m
LOCATION:1240 CS
DESCRIPTION:The class of datacenters known as “warehouse scale computers” (WSCs)  \n house large-scale data-intensive web services\, such as web search\, maps\,  \n social networking\, docs\, video sharing\, etc. Companies like Google\,  \n Microsoft\, Yahoo\, and Amazon spend tens to hundreds of millions of dollars to  \n construct and operate WSCs to provide these services. Maximizing the  \n efficiency of this class of computing devices reduces cost and has energy  \n implications for a greener planet. However\, WSC design and architecture  \n remains in its relative infancy. WSCs are built using commodity processor  \n architectures (Intel/AMD)\, and software components (Linux\, GCC\, JVM\, etc)  \n that have been engineered and optimized for traditional computing  \n environments and workloads\, such as those you would find in the  \n desktop/laptop/HPC environment. There are many characteristics\, assumptions\,  \n and requirements present in the WSC computing domain that impact design  \n decisions within these components. In this presentation\, we rethink how WSCs  \n are designed and architected\, identify sources of inefficiency\, and develop  \n solutions to improve WSCs\, with a particular focus on the interaction between  \n the application layer\, system-software stack\, and the underlying hardware  \n platform.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9148.field_date.0.95
SUMMARY:CS NEST Awards Ceremony
DTSTAMP:20130618T164412Z
DTSTART:20120413T203000Z
DTEND:20120413T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/cs-nest-awards-ceremony
LOCATION:CS 1240
DESCRIPTION:Awards ceremony for the third annual CS NEST Software Contest – an annual  \n software contest for UW Madison students to innovate and develop new software  \n technologies. Reception outside CS 1240 at 3:30. The presentations are all  \n day in CS 1240 from 9AM-2:50PM\, and the public is welcome to the  \n presentations too. See https://contest.cs.wisc.edu/ for more details.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9115.field_date.0.96
SUMMARY:Marshini Chetty: Faculty Candidate Talk: Marshini Chetty
DTSTAMP:20130618T164412Z
DTSTART:20120416T203000Z
DTEND:20120416T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/faculty-candidate-talk-marshini-chetty
LOCATION:1240 CS
DESCRIPTION:When Protocols Meet People: Interface Design For Home Networks Marshini  \n Chetty\, Georgia Insitute of Technology Abstract: We are increasingly  \n dependent on a working Internet connection and when the network is  \n functioning normally\, we hardly notice the underlying infrastructure. Yet\,  \n when the connection breaks\, we are reminded of how this complex ensemble of  \n protocols and bits can easily fail. Unfortunately\, although communication  \n networks were designed for robustness and scalability\, usability was never a  \n primary concern. As a result\, along with increased broadband Internet access  \n around the world\, the average person is faced with being a network operator  \n in the home. Users must deal with the underlying network protocols and arcane  \n technologies when all they really want is for the network to “just work”.  \n Designing an interface for a computer network therefore raises design  \n concerns that are common to many complex computer systems: How to expose the  \n right level of visibility and controls for user needs. Keeping the network as  \n a black box and limiting controls makes it harder for users to understand and  \n manage the system\, while too much visibility and control can make a system  \n overwhelming to operate\, understand\, and troubleshoot. In this talk\, I will  \n describe two systems that I have built to deploy in real user homes that  \n explore the parameters for visibility and controls in computer networks. My  \n first system\, Kermit\, explores questions of visibility and control around  \n network performance. I will briefly describe how I identified problems with  \n performance that home users face through studies of real-world households\,  \n how I addressed these issues with the design of a working system\, and the  \n results of an in situ deployment of Kermit. This study showed that users  \n respond positively to personalized visualizations of the network and controls  \n over performance that more closely match their needs for managing their  \n households. My second system\, uCap\, investigates issues of visibility and  \n control around network policies\, particularly under constrained Internet  \n conditions such as living with a bandwidth “cap” or limit on downloads  \n per a month. I will describe how users struggle with managing a bandwidth cap  \n because of a lack of visibility and control over how bandwidth is used by  \n various members of the household and different Internet applications. I will  \n then describe how these findings informed the design of uCap\, a system to  \n help households manage bandwidth caps by providing better visibility and  \n control over bandwidth usage in a home. Both of these systems demonstrate how  \n changing the parameters of visibility and control for complex computer  \n systems can improve the user experience. I will conclude with an agenda for  \n applying lessons from these studies more broadly across the design of other  \n complex infrastructure systems. Bio: Marshini Chetty is a postdoctoral  \n researcher in the College of Computing at Georgia Institute of Technology\,  \n where she recently graduated with her Ph.D. in Human-Centered Computing. Her  \n research in human computer interaction and ubiquitous computing focuses on  \n home networking\, sustainability\, and international development. She received  \n her Bachelors of Science and Masters of Science degrees in Computer Science  \n from the University of Cape Town. Marshini was awarded a Fulbright  \n Scholarship\, a Google Anita Borg Scholarship\, a GVU Foley Scholar award\, and  \n an Intel Ph.D. fellowship for her research. Her work has been featured in  \n technical blogs\, notably Slashdot\, Ars Technica\, Network World\, and  \n BoingBoing! and she received a CHI Best Paper Award in 2011.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9151.field_date.0.97
SUMMARY:Meg Walraed-Sullivan: Scalable Label Assignment in Data Center Networks
DTSTAMP:20130618T164412Z
DTSTART:20120417T191500Z
DTEND:20120417T201500Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/scalable-label-assignment-data-center-networks
LOCATION:CS 4310
DESCRIPTION:Modern data centers can consist of hundreds of thousands of servers and  \n millions of virtualized end hosts. A key challenge in providing scalable  \n communication in the data center is assigning identifiers\, or labels\, to  \n network elements and servers so that they can efficiently communicate and  \n perform cooperative tasks. The scale and complexity of a data center makes  \n the labeling problem unique in this environment and solutions often resort to  \n manual configuration that is costly\, time-consuming\, and error prone. In this  \n talk\, I will present ALIAS\, a distributed protocol for topology discovery and  \n label assignment in data center networks. ALIAS automates the assignment of  \n topologically meaningful addresses to the nodes in a data center\, enabling  \n scalable communication while significantly reducing the management burden of  \n manual configuration at scale. Meg is a PhD candidate in the Department of  \n Computer Science and Engineering at the University of California\, San Diego\,  \n working with Amin Vahdat and Keith Marzullo. Her research interests are  \n rooted in distributed systems and algorithms. She has recently applied this  \n investigation to the data center\, enabling scalable communication via  \n strategic label assignment and exploring the relationship between fault  \n tolerance and key properties of hierarchical topologies. Meg received a B.S.  \n and an M.Eng in Electrical and Computer Engineering from Cornell University  \n and will complete her PhD in Computer Science in the Summer of 2012.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9153.field_date.0.98
SUMMARY:Garret Suen: How Much Cow is in it? The Systems Biology of the Rumen  \n Ecosystem
DTSTAMP:20130618T164412Z
DTSTART:20120417T210000Z
DTEND:20120417T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/how-much-cow-it-systems-biology-rumen-ecosystem
LOCATION:Biotechnology Center...
DESCRIPTION:  Abstract: Ruminants like cows play an essential role in human agriculture.  \n This is particularly important in Wisconsin\, where milk and beef production  \n are critical to the state economy. Ruminants\, as dominant herbivores in North  \n America\, rely on microbial symbionts they harbour in their rumen to degrade  \n plant biomass to ferment nutrients for their host. Due to the domestication  \n efforts of humans\, ruminants have become one of the most efficient plant  \n biomass degrading and fermentation systems in nature. As a result\, the cow  \n has become a model system for biotechnological applications like the  \n production of advanced biofuels. Moreover\, the cow is now being revisited as  \n a model organism for understanding human health\, nutrition\, and disease\,  \n given the development of recent tools like a genome sequence and its genetic  \n tractability. Our own research is focused on characterizing and understanding  \n the complex microbial community found within the rumen. We have been using a  \n combination of 16S rRNA sequencing\, whole-genome sequencing\, and RNA-seq to  \n characterize what specific microbes are found in the rumen and their  \n potential functional roles. In particular\, we are interested in how these  \n microbes influence milk production and the mechanisms through which specific  \n bacteria ferment plant biomass into short chain fatty acids. We show that  \n there are differences in the ruminal microbial community between high and low  \n milk producing cows\, particularly in those bacteria that degrade cellulose.  \n Whole-genome sequencing of two of these cellulolytic bacteria\, Fibrobacter  \n succinogenes and Rumincococus albus\, reveal uniquely different approaches to  \n cellulose degradation. Morever\, transcript sequencing of R. albus on  \n cellulose and cellobiose surprisingly revealed that the most highly expressed  \n genes include the tryptophan biosynthesis operon. This has led us to  \n determine that tryptophan is found in greater abundance in proteins  \n associated with cellulose degradation\, when compared to other genes in the  \n genome. By applying these approaches we are beginning to understand not only  \n how ruminal bacterial communities are associated with host production\, but  \n also how specific members of the community are contributing to the conversion  \n of feed into host-usable nutrients.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9159.field_date.0.99
SUMMARY:Nick Feamster: The Battle for Control of Online Communications
DTSTAMP:20130618T164412Z
DTSTART:20120419T210000Z
DTEND:20120419T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/battle-control-online-communications
LOCATION:CS 1240
DESCRIPTION:  The Internet offers users many opportunities for communicating and  \n exchanging ideas\, but abuse\, censorship\, and the manipulation of Internet  \n traffic have put free and open communication at risk. Recent estimates  \n suggest that spam constitutes about 95% of all email traffic\; hundreds of  \n thousands of online scam domains emerge every day\; online social networks may  \n be used to spread propaganda\; and more than 60 countries around the world  \n censor Internet traffic. In this talk\, I will present approaches that we have  \n developed to preserve free and open communication on the Internet in the face  \n of these threats. First\, I will describe the threat of message abuse (e.g.\,  \n spam) and describe methods we have developed for mitigating it. I will  \n briefly discuss a 13-month study of the network-level behavior of spammers\,  \n and present SNARE\, a spam filtering system we developed that classifies email  \n messages based on the network-level traffic characteristics of the email  \n messages\, rather than their contents. Next\, I will turn to information  \n censorship\, and describe Collage\, a system that circumvents censorship  \n without arousing the suspicion of the censor. Finally\, I will discuss the  \n various forms of information manipulation\, including the spread of propaganda  \n in social networks and online "filter bubbles". Although it is difficult to  \n prevent all forms of manipulation\, our goal is to make it more transparent to  \n users. Towards this goal\, I will describe my broader research agenda and  \n plans\, which aim to improve Internet transparency for aspects of Internet  \n communication ranging from network performance to social media to search  \n results using the aggregation of data from a wide variety of vantage points.  \n Bio: Nick Feamster is an associate professor in the College of Computing at  \n Georgia Tech. He received his Ph.D. in Computer science from MIT in 2005\, and  \n his S.B. and M.Eng. degrees in Electrical Engineering and Computer Science  \n from MIT in 2000 and 2001\, respectively. His research focuses on many aspects  \n of computer networking and networked systems\, including the design\,  \n measurement\, and analysis of network routing protocols\, network operations  \n and security\, and anonymous communication systems. In December 2008\, he  \n received the Presidential Early Career Award for Scientists and Engineers  \n (PECASE) for his contributions to cybersecurity\, notably spam filtering. His  \n honors include the Technology Review 35 "Top Young Innovators Under 35"  \n award\, a Sloan Research Fellowship\, the NSF CAREER award\, the IBM Faculty  \n Fellowship\, and award papers at SIGCOMM 2006 (network-level behavior of  \n spammers)\, the NSDI 2005 conference (fault detection in router  \n configuration)\, Usenix Security 2002 (circumventing web censorship using  \n Infranet)\, and Usenix Security 2001 (web cookie analysis).
END:VEVENT
BEGIN:VEVENT
UID:calendar.9160.field_date.0.100
SUMMARY:David Du: A New Era for the Convergence of Network Centric and Data Centric  \n Computing
DTSTAMP:20130618T164412Z
DTSTART:20120420T170000Z
DTEND:20120420T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/new-era-convergence-network-centric-and-data-centric-computing
LOCATION:2310 CS
DESCRIPTION:The Internet today has grown to an enormously large scale. Devices large and  \n small are connected globally from anywhere on the earth. With the rapid  \n advancement of technology\, we now also have cheap and small devices with high  \n computing power and large storage capacity. These devices are designed to  \n improve our daily life by monitoring our environment\, collecting critical  \n data\, and executing special instructions. These devices have gradually become  \n a dominant part of our Internet. Many imaging\, audio and video data are  \n converted from analog to digital and digital data are generated at an  \n alarming rate. As a result\, unprecedented amount of data are available. How  \n to manage and look for the desired information becomes a great challenge. How  \n to preserve these data becomes a crisis. At the same time\, many emerging  \n applications like service-oriented\, security and real-time demand much better  \n support than the current Internet can offer. In this talk\, we will present a  \n vision of content addressable future Internet. What are the essential changes  \n in data representation\, information retrieval\, storage systems and networking  \n design will be discussed. We believe an object-oriented intelligent storage  \n is an essential part of the solution to this new computing and communication  \n environment. We will also present a number of research projects that are  \n currently under investigation in our NSF I/UCRC Center on Intelligent  \n Storage. These projects include data deduplication\, long-term data  \n preservation\, data center power management\, and flash memory based solid  \n state drives. Bio: David Hung-Chang Du: Dr. Du is currently the Qwest Chair  \n Professor of Computer Science and Engineering at University of Minnesota\,  \n Minneapolis. He has served as a Program Director (IPA) at National Science  \n Foundation (NSF) CISE/CNS Division from 2006 to 2008. At NSF\, he was  \n responsible for NeTS (networking research cluster) NOSS (Networks of Sensor  \n Systems) Program and worked on Cyber Trust (Internet Security) Program. He is  \n also the Director of a NSF I/UCRC Center on Intelligent Storage. Dr. Du  \n received a B.S. degree from National Tsing Hua Univeristy in 1974\, an M.S.  \n and Ph.D. degree from University of Washington (Seattle) in 1980 and 1981  \n respectively. He joined University of Minnesota as a faculty since 1981. Dr.  \n Du has a wide range of research expertise including multimedia computing\,  \n mass storage systems\, high-speed networking\, sensor networks\, cyber security\,  \n high-performance file systems and I/O\, database design\, and CAD for VLSI  \n circuits. He has authored and co-authored over 230 technical papers including  \n 110 referred journal publications in these research areas. He has graduated  \n 51 Ph.D. and 80 M.S. students in the last 30 years. Dr. Du is an IEEE Fellow  \n (since 1998) and a Fellow of the Minnesota Supercomputer Institute.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9165.field_date.0.101
SUMMARY:Benjamin Shapiro\, Research Associate: Improving Education through Modeling  \n Learning Interactions with Digital Media
DTSTAMP:20130618T164412Z
DTSTART:20120423T210000Z
DTEND:20120423T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/improving-education-through-modeling-learning-interactions-digital-media
LOCATION:Wisconsin Institute for...
DESCRIPTION:My research focuses on how to design learning environments that a) immerse  \n participants in the authentic practices and ways of thinking of engineering\,  \n arts\, and sciences\; b) dynamically respond to learners' individual and  \n collective skill development\; c) actively support learning facilitation by  \n teachers\, parents\, and other caregivers\; and d) certify learning  \n accomplishments through valid measures of competence derived from analyses of  \n learners' participation in these environments. The Educational Research group  \n of the Wisconsin Institutes for Discovery\, in which I am a postdoc\, is  \n pursuing these aims through the creation of educational video games. Though  \n we have world-leading expertise in the design of engaging and authentic games  \n about science\, we are just beginning to explore how to imbue these  \n environments with the responsive dynamism that goals b\, c\, d (above) require.  \n Assessment of learning over time is at the heart of the challenges we now  \n face in realizing them. In this talk\, I'll present some of the core theories  \n of learning that guide our work\, discuss essential challenges to their  \n application in practice\, and give an overview of our work on Trails Forward\,  \n a game we are creating in collaboration with Michael Ferris. Then\, I will  \n explain how the essence of the assessment and responsive design challenge we  \n now face is essentially one of modeling time series of interactions between  \n learners and game\, as well as between learners and one-another\, as evidence  \n for learning\, illustrating how the learning (and modeling) goals we have are  \n fundamentally different than those that the educational data mining  \n literature has tackled thus far. I will conclude with an invitation for  \n collaboration between the UW/WID Education Research and Optimization  \n communities.<?xml:namespace prefix = o />
END:VEVENT
BEGIN:VEVENT
UID:calendar.9163.field_date.0.102
SUMMARY:Deborah Chasman: Inferring Host Subnetworks Involved in Viral Replication
DTSTAMP:20130618T164412Z
DTSTART:20120424T210000Z
DTEND:20120424T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/inferring-host-subnetworks-involved-viral-replication
LOCATION:Biotechnology Center...
DESCRIPTION:  Abstract: Understanding the interactions that occur between viruses and  \n their hosts is important for controlling the impact of viruses on human  \n health. Systematic\, genome-wide loss-of-function experiments can be used to  \n identify host factors that directly or indirectly facilitate or inhibit the  \n replication of a virus in a host cell. We present a computational approach  \n that uses an integer linear program to infer the intracellular pathways  \n through which these host factors modulate viral replication. The input is a  \n set of viral phenotypes observed in single-host-gene mutants and a background  \n network consisting of a variety of host cell intracellular interactions. The  \n output is an ensemble of subnetworks that provides a consistent explanation  \n for each gene’s role in viral replication\, predicts which unassayed host  \n factors modulate the virus\, and predicts which host factors are the most  \n direct interfaces with a viral component. The value of these inferred  \n subnetworks is that they can be used to guide further experimentation toward  \n uncovering and validating the mechanisms of host-virus interactions. We  \n analyze data from experiments screening the yeast genome for genes modulating  \n the replication of two RNA viruses. To evaluate our method\, we conduct a  \n cross-validation experiment in which we predict whether held-aside test genes  \n have an effect on viral replication. Our method is able to make these  \n predictions with accuracy greater than or equal to several baseline methods  \n that do not posit mechanistic pathways. As an additional evaluation\, we use  \n our approach to predict which unassayed host genes are likely to be involved  \n in viral replication. Multiple predictions are supported by recent  \n independent experimental data. (This work is a collaboration with Brandi  \n Gancarz\, Linhui Hao\, Michael Ferris\, Paul Ahlquist\, and Mark Craven.)
END:VEVENT
BEGIN:VEVENT
UID:calendar.9166.field_date.0.103
SUMMARY:Andrew Bernat: Abstract\, Safe\, Timely\, and Efficient Binary Modification
DTSTAMP:20130618T164412Z
DTSTART:20120425T144500Z
DTEND:20120425T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/abstract-safe-timely-and-efficient-binary-modification
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: Bart Miller (advisor)\;Tom Reps\, Jeff Hollingsworth\, Mike Swift\,  \n Paul Wilson                            
END:VEVENT
BEGIN:VEVENT
UID:calendar.9172.field_date.0.104
SUMMARY:Professor Shuchi Chawla: Professors & Pizza - Algorithms
DTSTAMP:20130618T164412Z
DTSTART:20120430T203000Z
DTEND:20120430T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/professors-pizza-algorithms
LOCATION:CS1240
DESCRIPTION:Not sure which CS class to take next? Come find out why you should take Intro  \n to Algorithms (CS577) while eating FREE PIZZA! Time: 3:30pm Date: Monday\,  \n April 30th Location: CS1240 Abstract: Algorithms are recipes for problem  \n solving. A solid algorithmic arsenal forms the foundation for much of  \n computer science and is a much sought-after skill in CS job interviews. This  \n course introduces basic techniques for the design and analysis of algorithms\,  \n with applications in AI\, computational biology\, network protocols\, and  \n optimization. You will learn how to multiply faster than the grade school  \n approach\, why no sorting algorithm can beat mergesort/quicksort\, and what the  \n P vs. NP question is all about. Speaker information: Shuchi Chawla is an  \n assistant professor of Computer Sciences at the University of Wisconsin\,  \n Madison. Her recent research involves designing algorithms for optimization  \n problems arising in economic settings. She is the recipient of an NSF Career  \n Award and a Sloan Foundation fellowship.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9174.field_date.0.105
SUMMARY:Kevin Roundy: Kevin Roundy Ph.D. Final Oral Defense
DTSTAMP:20130618T164412Z
DTSTART:20120501T140000Z
DTEND:20120501T163000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/kevin-roundy-phd-final-oral-defense
LOCATION:3310 Computer Sciences
DESCRIPTION:Title: Hybrid Analysis and Control of Malicious Code Committee:Bart Miller  \n (advisor)\; Tom Ristenpart\, Somesh Jha\, Shan Lu and Nigel Boston
END:VEVENT
BEGIN:VEVENT
UID:calendar.9171.field_date.0.106
SUMMARY:Brian Teague: Distributions and DNA: A Hidden Markov Model for Optical  \n Mapping Data
DTSTAMP:20130618T164412Z
DTSTART:20120501T210000Z
DTEND:20120501T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/distributions-and-dna-hidden-markov-model-optical-mapping-data
LOCATION:Biotechnology Center...
DESCRIPTION:Abstract: Optical Mapping is a unique platform for analyzing genomes: it uses  \n measurements of single molecules of DNA to infer genome structure\, which  \n complements genome sequence to yield biological insight. Information on a  \n genome's structure is useful in a wide variety of contexts\, including aiding  \n in sequence assembly\; understanding normal human genetic variation\; and  \n probing cancer genomes in search of new therapeutic targets. Oncoming  \n advances in DNA enzymology\, molecule presentation and image analysis herald  \n dramatic improvements in Optical Mapping's speed and resolution\, promising  \n deeper understanding of the biology of genomes. Alongside improvements in  \n sample preparation and data collection\, advances in algorithms and analyses  \n allow us to probe these single molecule data sets in new ways\, asking  \n questions that were previously inaccessible. This seminar will describe the  \n development of a hidden Markov model (HMM) for Optical Mapping data.  \n Developed at Bell Labs in the 1970s for voice recognition\, HMMs have found  \n use in a variety of bioinformatics endeavours including gene finding\, copy  \n number analysis\, secondary structure prediction and multiple sequence  \n alignment. Their success is founded on their ability to perform inference on  \n systems whose internal state is *hidden* by noisy or incomplete data\,  \n resulting in algorithms that are fast\, accurate and well-grounded in theory.  \n They work particularly well when paired with large data sets whose error  \n processes are well-characterized. Happily\, the data sets produced by the  \n Optical Mapping platform fit this description: we want to infer properties of  \n the genome based on a large ensemble of single-molecule observations whose  \n generation is subject to a number of well-characterized error processes. The  \n best-studied problems on hidden Markov models (evaluation\, decoding\, and  \n learning) translate directly to common tasks in analyzing Optical Mapping  \n data\, and provide a jumping-off point for solving more difficult (but more  \n interesting!) problems including map refinement and haplotype discernment.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9168.field_date.0.107
SUMMARY:Florentina Popovici: WACM Speaker Series: Florentina Popovici From Google
DTSTAMP:20130618T164412Z
DTSTART:20120502T210000Z
DTEND:20120502T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/wacm-speaker-series-florentina-popovici-google
LOCATION:CS 3310
DESCRIPTION:Florentina Popovici currently works at Google\, where she focuses  \n on distributed storage. She received her doctorate from the Computer  \n Sciences Department at University of Wisconsin-Madison under the direction  \n of Andrea Arpaci-Dusseau and Remzi Arpaci-Dusseau. One of her fond memories  \n from those times is setting a Datamation world sort record\, which will never  \n be claimed by anybody else. She is a two-time winner of the IBM  \n PhD fellowship award. She enjoys the analysis\, design\, and implementation  \n of large scale distributed storage systems. At Google she worked on  \n the Google File System\, and currently she analyzes and helps design the  \n storage system for availability and performance.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9169.field_date.0.108
SUMMARY:Romit Roy Choudhury: Smartphone Positioning Systems
DTSTAMP:20130618T164412Z
DTSTART:20120503T210000Z
DTEND:20120503T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/smartphone-positioning-systems
LOCATION:CS 1240
DESCRIPTION:The sudden boom in the smartphone ecosystem has caught various branches of  \n technology unprepared -- one of these branches is localization. An escalating  \n number of location-based apps are demanding tailor-made solutions\; the  \n futuristic ones are even broadening the core notion of ``location''. For  \n instance\, the advertisement industry is asking for semantic localization  \n services\, wherein the device's location is expressed as ``Starbucks'' or  \n ``Wal-Mart''. Museums intend to precisely identify the painting a visitor is  \n facing\, to be able to offer information about it. Reminder apps are calling  \n for continuous localization even though the phone runs on a limited energy  \n budget. Social apps aim to present walking directions within a shopping mall\,  \n so Alice can find a way to reach Bob. Finally\, augmented reality apps are  \n aspiring for a technology that localizes visible objects -- the ability to  \n look at a distant building through the phone and obtain its location.  \n Clearly\, GPS was not designed to serve this wide spectrum of  \n application-specific demands. The landscape of localization needs to be  \n revamped against the backdrop of emerging constraints and opportunities. This  \n talk will describe our efforts in this direction -- the multiple failures\,  \n and a recent promise of success. Biography Romit Roy Choudhury is an  \n Associate Professor of ECE and CS at Duke University. He joined Duke in Fall  \n 2006\, after completing his PhD from UIUC. His research interests are in  \n wireless protocol design mainly at the PHY/MAC layer\, and in mobile computing  \n at the application layer. He received the NSF CAREER Award in January 2008.  \n Visit Romit's Systems Networking Research Group (SyNRG)\, at  \n http://synrg.ee.duke.edu
END:VEVENT
BEGIN:VEVENT
UID:calendar.9177.field_date.0.109
SUMMARY:Nisha Talagala: The Implications of Non-Volatile Memory on Software  \n Architectures
DTSTAMP:20130618T164412Z
DTSTART:20120507T160000Z
DTEND:20120507T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/implications-non-volatile-memory-software-architectures
LOCATION:4310 CS
DESCRIPTION:Flash based non volatile memory is revolutionizing data center architectures\,  \n improving application performance by bridging the gap between DRAM and disk.  \n Future non volatile memories promise performance even closer to DRAM. While  \n flash adoption in industry started as disk replacement\, the past several  \n years have seen data center architectures change to take advantage of flash  \n as a new memory tier in both servers and storage. This talk covers the  \n implications of nonvolatile memory on software. We describe the stresses that  \n non volatile memory places on existing application and OS designs\, and  \n illustrate optimizations to exploit flash as a new memory tier. Until the  \n introduction of flash\, there has been no compelling reason to change the  \n existing operating system storage stack. We will describe the technologies  \n contained in the upcoming Fusion-io Software Developer Kit (ioMemory SDK)  \n that allow applications to leverage the native capabilities of non-volatile  \n memory as both an I/O device and a memory device. The technologies described  \n will include new I/O based APIs and libraries to leverage the ioMemory  \n Virtual Storage Layer\, as well as features for extending DRAM into flash for  \n cost and power reduction. Finally\, we describe Auto-Commit-Memory\, a new  \n persistent memory type that will allow applications to combine the benefits  \n of persistence with programming semantics and performance levels normally  \n associated with DRAM . Bio: Nisha Talagala is Lead Architect at Fusion-io\,  \n where she works on innovation in non volatile memory technologies and  \n applications. Nisha has more than 10 years of expertise in software  \n development\, distributed systems\, storage\, I/O solutions\, and non-volatile  \n memory. She has worked as technology lead for server flash at Intel - where  \n she led server platform non volatile memory technology development and  \n partnerships. Prior to Intel\, Nisha was the CTO of Gear6\, where she developed  \n clustered computing caches for high performance I/O environments. Nisha also  \n served at Sun Microsystems\, where she developed storage and I/O solutions and  \n worked on file systems. Nisha earned her PhD at UC Berkeley where she did  \n research on clusters and distributed storage. Nisha hold more than 30 patents  \n in distributed systems\, networking\, storage\, performance and non-volatile  \n memory.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9190.field_date.0.110
SUMMARY:Balasubramanian Sivan : Prior Independent Optimization
DTSTAMP:20130618T164412Z
DTSTART:20120507T180000Z
DTEND:20120507T200000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/prior-independent-optimization
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: Shuchi Chawla (advisor)\, Eric Bach\, Dieter van Melkebeek\, Marzena  \n Rostek  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9145.field_date.0.111
SUMMARY:Kurt M. Anstreicher\, Professor: Second-Order-Cone Constraints for Extended  \n Trust-Region Subproblems
DTSTAMP:20130618T164412Z
DTSTART:20120507T210000Z
DTEND:20120507T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/second-order-cone-constraints-extended-trust-region-subproblems
LOCATION:WID\, Room 3280b\, (3rd...
DESCRIPTION:The classical trust-region subproblem (TRS) minimizes a nonconvex quadratic  \n objective over the unit ball. We consider extensions of TRS having additional  \n constraints. It is known that TRS\, and the extension of TRS that adds a  \n single linear inequality\, both admit convex programming representations. We  \n show that when two parallel linear inequalities are added to TRS\, the  \n resulting nonconvex problem has an exact convex representation as a  \n semidefinite programming (SDP) problem with additional linear and  \n second-order-cone constraints. For the case where an additional ellipsoidal  \n constraint is added to TRS\, resulting in the well-known “two trust-region  \n subproblem” (TTRS)\, we describe a new relaxation including  \n second-order-cone constraints that significantly strengthens the usual SDP  \n relaxation. Numerical experiments show that the strengthened relaxation  \n provides an exact solution of TTRS in most instances\, although the  \n theoretical complexity of TTRS remains an open problem. This is joint work  \n with Sam Burer. <?xml:namespace prefix = o />  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9191.field_date.0.112
SUMMARY:Arkaprava Basu: Reducing Memory Management Waste
DTSTAMP:20130618T164412Z
DTSTART:20120508T180000Z
DTEND:20120508T200000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/reducing-memory-management-waste
LOCATION:4310
DESCRIPTION:Committee: Mark Hill (advisor)\, Michael Swift (advisor)\, Remzi  \n Arpaci-Dusseau\, David Wood\, Mikko Lipasti
END:VEVENT
BEGIN:VEVENT
UID:calendar.9189.field_date.0.113
SUMMARY:Project Demo Day for CS638 [Software Engineering]
DTSTAMP:20130618T164412Z
DTSTART:20120508T193000Z
DTEND:20120508T204500Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/project-demo-day-cs638-software-engineering
LOCATION:1st Floor Unit III...
DESCRIPTION:CS638 [Software Engineering] students work together in large semester project  \n teams. Visit us on the first floor of Unit III to see live demos of: Android  \n Text Message Search\, proposed by Sarah Gilliland\, developed by Benjamin  \n Frisch\, Kong Yang\, Sachin Ravi\, Sarah Gilliland\, and Theodora Hinkle  \n Artificial Intelligence Agent for StarCraft:Broodwar\, proposed by Michael  \n Starr\, developed by Andrew Reichert\, Brian Ploeckelman\, Diansheng Liu\, James  \n Loethen\, Matthew Linson\, and Michael Starr GoCuro Mobile Material Tracker\,  \n proposed by John Terrill\, developed by Borui Wang\, Jared Lutteke\, John  \n Terrill\, Le Yu\, and Yim Yoong Madison Student Housing Finder\, proposed by  \n Matt Wysocki\, developed by Adam Eggum\, Cory Romdenne\, Guilherme Santos Galvao  \n Baptista\, Justin Smith\, and Matt Wysocki Org Organizer\, proposed by Joe  \n Green\, developed by Brennan Payne\, David Mayer\, Drew Smith\, James Schindler\,  \n and Joe Green These brave student test pilots were the first to go through  \n our brand new software engineering course. Please join us to thank them and  \n to celebrate their accomplishments!
END:VEVENT
BEGIN:VEVENT
UID:calendar.9198.field_date.0.114
SUMMARY:Distributed Systems (CS 739) Project Poster Session
DTSTAMP:20130618T164412Z
DTSTART:20120510T160000Z
DTEND:20120510T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/distributed-systems-cs-739-project-poster-session
LOCATION:2310
DESCRIPTION: CS 739 - Distributed Systems - project poster session. Students in 739 will  \n present their cloud computing projects on posters. Topics include distributed  \n file systems and cloud applications.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9195.field_date.0.115
SUMMARY:Matthew Anderson: Advancing Algebraic and Logical Approaches to Circuit Lower  \n Bounds
DTSTAMP:20130618T164412Z
DTSTART:20120510T193000Z
DTEND:20120510T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/advancing-algebraic-and-logical-approaches-circuit-lower-bounds
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: Dieter van Melkeebeek (advisor)\; Eric\, Jin-Yi Cai\, Shuchi Chawla\,  \n Lance Fortnow\, Donald Passman
END:VEVENT
BEGIN:VEVENT
UID:calendar.9196.field_date.0.116
SUMMARY:Michael J. Brim: Control and Inspection of Distributed Process Groups at  \n Extreme Scale via Group File Semantics
DTSTAMP:20130618T164412Z
DTSTART:20120511T143000Z
DTEND:20120511T163000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/control-and-inspection-distributed-process-groups-extreme-scale-group-file-semantics
LOCATION:4310
DESCRIPTION:Committe: Barton Miller (advisor)\, Remzi Arpaci-Dusseau\, Michael Swift\,  \n Benjamin Liblit\, Rafael Lazimy
END:VEVENT
BEGIN:VEVENT
UID:calendar.9203.field_date.0.117
SUMMARY:Ian Rae: Two Fish Out of Water? Native Full-text Search in RDBMSs and a  \n Relational Interface for Hadoop
DTSTAMP:20130618T164412Z
DTSTART:20120514T140000Z
DTEND:20120514T160000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/two-fish-out-water-native-full-text-search-rdbmss-and-relational-interface-hadoop
LOCATION:3310 Computer Sciences
DESCRIPTION:Commmittee: Jeff Naughton\, David Dewitt\, Jignessh Patel\, Chris Re
END:VEVENT
BEGIN:VEVENT
UID:calendar.9197.field_date.0.118
SUMMARY:Shan-Hsiang Shen: Building Blocks for Content-aware Communication
DTSTAMP:20130618T164412Z
DTSTART:20120514T190000Z
DTEND:20120514T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/building-blocks-content-aware-communication
LOCATION:3310
DESCRIPTION:Committee: Aditya Akella (advisor) \, Suman Banerjee\, Shan Lu
END:VEVENT
BEGIN:VEVENT
UID:calendar.9204.field_date.0.119
SUMMARY:Seeun William Umboh: Algorithms for Resource Allocation and Scheduling
DTSTAMP:20130618T164412Z
DTSTART:20120514T193000Z
DTEND:20120514T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/algorithms-resource-allocation-and-scheduling
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: Suchi Chawla (advisor)\, Eric Bach (advisor)\, Dieter van Melkebeek
END:VEVENT
BEGIN:VEVENT
UID:calendar.9192.field_date.0.120
SUMMARY:CS252 Demos: CS252 Demos (Intro to Computer Engineering)
DTSTAMP:20130618T164412Z
DTSTART:20120514T210000Z
DTEND:20120514T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/cs252-demos-intro-computer-engineering
LOCATION:CS Lobby
DESCRIPTION:Come see the projects our undergraduate freshman have completed in CS 252.  \n Using the arduino hardware board and sensors\, the students have programmed  \n an interesting set of projects for 252.  There are 5 types of projects with  \n a total of about 15 teams.  They are: 1) Show a twitter feed on an LCD 2)  \n Automatically navigate a robot through a maze 3) Obstacle avoiding robot 4)  \n Wireless 2-player touch-sensitive tic-tac-toe game 5) Physical Angry birds -  \n throw an angry bird at target
END:VEVENT
BEGIN:VEVENT
UID:calendar.9170.field_date.0.121
SUMMARY:Junming Sui: Learning from Bullying Traces in Social Media
DTSTAMP:20130618T164412Z
DTSTART:20120517T210000Z
DTEND:20120517T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/learning-bullying-traces-social-media
LOCATION:CS 3310
DESCRIPTION:This is a practice talk for the 2012 Conference of the North American Chapter  \n of the Association for Computational Linguistics: Human Language Technologies  \n (NAACL-HLT) Abstract: We introduce the social study of bullying to the NLP  \n community. Bullying\, in both physical and cyber worlds (the latter known as  \n cyberbullying)\, has been recognized as a serious national health issue among  \n adolescents. However\, previous social studies of bullying are handicapped by  \n data scarcity\, while the few computational studies narrowly restrict  \n themselves to cyberbullying which accounts for only a small fraction of all  \n bullying episodes. Our main contribution is to present evidence that social  \n media\, with appropriate natural language processing techniques\, can be a  \n valuable and abundant data source for the study of bullying in both worlds.  \n We identify several key problems in using such data sources and formulate  \n them as NLP tasks\, including text classification\, role labeling\, sentiment  \n analysis\, and topic modeling. Since this is an introductory paper\, we present  \n baseline results on these tasks using off-the-shelf NLP solutions\, and  \n encourage the NLP community to contribute better models in the future.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9175.field_date.0.122
SUMMARY:Ph.D. Graduation Reception
DTSTAMP:20130618T164412Z
DTSTART:20120518T200000Z
DTEND:20120518T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/phd-graduation-reception
LOCATION:WID Forum
DESCRIPTION:Reception for Ph.D. graduates\, thier invited guests\, and faculty.  Portraits  \n for graduates will be taken by Perry Kivolowitz from 1:30-3:00 in the Forum  \n prior to the reception.  RSVP required. Ph.D. commencement at Kohl Center  \n will follow CS reception.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9176.field_date.0.123
SUMMARY:CS Reception for MS and Undergraduates Spring Graduates
DTSTAMP:20130618T164412Z
DTSTART:20120520T170000Z
DTEND:20120520T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/cs-reception-ms-and-undergraduates-spring-graduates
LOCATION:Fluno Center
DESCRIPTION:Reception for graduating undergraduates\, M.S. graduates.\, their invited  \n guests and faculty. RSVP required. Sunday Brunch Buffet will be served.   \n Perry Kivolovitz will be onsite to take portrait photos of graduates from  \n 11:30- 2 pm.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9207.field_date.0.124
SUMMARY:Srinath Sridharan: Adaptive Execution of General Purpose Parallel Programs
DTSTAMP:20130618T164412Z
DTSTART:20120521T150000Z
DTEND:20120521T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/adaptive-execution-general-purpose-parallel-programs
LOCATION:4310 CS
DESCRIPTION:Committee: Gurindar Sohi (advisor)\; David Wood\, Karu Sankaralingam\, Remzi  \n Arpaci-Dusseau\, Mikko Lipasti
END:VEVENT
BEGIN:VEVENT
UID:calendar.9206.field_date.0.125
SUMMARY:Constantine Dovrolis: CE Seminar: Hourglass-like networks: from protocol  \n stacks to developing embryos
DTSTAMP:20130618T164412Z
DTSTART:20120521T193000Z
DTEND:20120521T203000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/ce-seminar-hourglass-networks-protocol-stacks-developing-embryos
LOCATION:Room 4610\, Engineering...
DESCRIPTION:Abstract: The architecture of both technological and natural complex systems  \n often exhibits hierarchical modularity: simpler modules are used in the  \n construction of increasingly more complex modules. These architectures can be  \n modeled as ``layered design networks''\, i.e.\, directed\, acyclic and layered  \n graphs in which the simplest modules appear at the bottom layer and the most  \n complex modules are produced at the highest layer. Such layered design  \n networks are often subject to an evolutionary process in which modules at the  \n same layer and serving almost the same function compete and potentially  \n replace each other. We propose a model of this evolutionary layered design  \n process. The model shows that the structure of a layered design network  \n naturally takes the shape of an hourglass\, with the modules at the waist of  \n the hourglass being the oldest and most resistant to change. This talk will  \n focus on the application of this model in the context of networking protocol  \n stacks and biological development. Biography: Dr. Constantine Dovrolis is an  \n Associate Professor at the College of Computing of the Georgia Institute of  \n Technology. He received the Computer Engineering degree from the Technical  \n University of Crete in 1995\, the M.S. degree from the University of Rochester  \n in 1996\, and the Ph.D. degree from the University of Wisconsin-Madison in  \n 2000. He joined Georgia Tech in August 2002\, after serving at the faculty of  \n the University of Delaware for about two years. He has held visiting  \n positions at Thomson Research in Paris\, Simula Research in Oslo\, and FORTH in  \n Crete. His current research focuses on the evolution of the Internet\,  \n Internet economics\, and on applications of network measurement. He is also  \n interested in cross-disciplinary applications of network science in biology\,  \n climate science and neuroscience. Dr. Dovrolis has been an editor for the  \n IEEE/ACM Transactions on Networking\,the ACM Communications Review (CCR)\, and  \n he served as the Program co-Chair for PAM'05\, IMC'07\, CoNEXT'11\, and as the  \n General Chair for HotNets'07. He received the National Science Foundation  \n CAREER Award in 2003.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9205.field_date.0.126
SUMMARY:Linhai Song: Understanding and Detecting Real-World Performance Bugs
DTSTAMP:20130618T164412Z
DTSTART:20120524T180000Z
DTEND:20120524T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/understanding-and-detecting-real-world-performance-bugs
LOCATION:3310
DESCRIPTION:Developers frequently use inefficient code sequences that could be fixed by  \n simple patches. These inefficient code sequences can cause significant  \n performance degradation and resource waste\, referred to as performance bugs.  \n Meager increases in single threaded performance in the multi-core era and  \n increasing emphasis on energy efficiency call for more effort in tackling  \n performance bugs.  This paper conducts a comprehensive study of 109  \n real-world performance bugs that are randomly sampled from five  \n representative software suites (Apache\, Chrome\, GCC\, Mozilla\, and MySQL). The  \n findings of this study provide guidance for future work to avoid\, expose\,  \n detect\, and fix performance bugs. Guided by our characteristics study\,  \n efficiency rules are extracted from 25 patches and are used to detect  \n performance bugs. 332 previously unknown performance problems are found in  \n the latest versions of MySQL\, Apache\, and Mozilla applications\, including 219  \n performance problems found by applying rules across applications.  This is a  \n practice talk for PLDI. Candy\, bagels\, coffee and tea will be served. 
END:VEVENT
BEGIN:VEVENT
UID:calendar.9212.field_date.0.127
SUMMARY:Mariyam Mirza: A Machine Learning Based Approach to Problems in Computer  \n Network Measurement and Performance Analysis
DTSTAMP:20130618T164412Z
DTSTART:20120525T150000Z
DTEND:20120525T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/machine-learning-based-approach-problems-computer-network-measurement-and-performance-analysis
DESCRIPTION:Committee: Paaul Barford (advisor)\, Jerry Zhu\, Suman Banerjee\, Aditya Akella\,  \n Rob Novak  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9213.field_date.0.128
SUMMARY:Mariyam Mirza: A Machine Learning Based Approach to Problems in Computer  \n Network Measurement and Performance Analysis
DTSTAMP:20130618T164412Z
DTSTART:20120525T150000Z
DTEND:20120525T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/machine-learning-based-approach-problems-computer-network-measurement-and-performance-analys-0
DESCRIPTION:Committee: Paaul Barford (advisor)\, Jerry Zhu\, Suman Banerjee\, Aditya Akella\,  \n Rob Novak  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9208.field_date.0.129
SUMMARY:Brandon Michael Smith: Face Alignment with Semantic Segmentation for  \n Interactive Face Image Retrieval
DTSTAMP:20130618T164412Z
DTSTART:20120525T183000Z
DTEND:20120525T203000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/face-alignment-semantic-segmentation-interactive-face-image-retrieval
LOCATION:3310
DESCRIPTION:Committee: Li Zhang (advisor)\, Charles Dyer\, Michael Gleicher
END:VEVENT
BEGIN:VEVENT
UID:calendar.9217.field_date.0.130
SUMMARY:Marc de Kruijf: Static Analysis and Compiler Design for Idempotent Processing
DTSTAMP:20130618T164412Z
DTSTART:20120604T210000Z
DTEND:20120604T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/static-analysis-and-compiler-design-idempotent-processing
LOCATION:CS 4310
DESCRIPTION:Recovery functionality has many applications in computing systems\, from  \n speculation recovery in modern microprocessors to fault recovery in  \n high-reliability systems. Modern systems commonly recover using checkpoints.  \n However\, checkpoints introduce overheads\, add complexity\, and often save more  \n state than necessary.     In this talk\, I will present a novel compiler  \n technique to recover program state without the overheads of explicit  \n checkpoints. The technique breaks programs into idempotent regions—regions  \n that can be freely re-executed—which allows recovery without checkpointed  \n state.  Leveraging the property of idempotence\, recovery can be obtained by  \n simple re-execution. We develop static analysis techniques to construct these  \n regions and demonstrate low overheads and large region sizes for an  \n LLVM-based implementation. Across a set of diverse benchmark suites\, we  \n construct idempotent regions close in size to those that could be obtained  \n with perfect runtime information. Although the resulting code runs more  \n slowly\, typical performance overheads are in the range of just 2-12%.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9220.field_date.0.131
SUMMARY:Arka Basu: Reducing Memory Reference Energy with Opportunistic Virtual  \n Caching
DTSTAMP:20130618T164412Z
DTSTART:20120605T210000Z
DTEND:20120605T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/reducing-memory-reference-energy-opportunistic-virtual-caching
LOCATION:CS 1221
DESCRIPTION:Most modern cores perform a highly-associative transaction look aside buffer  \n (TLB) lookup on every memory access. These designs often hide the TLB lookup  \n latency by overlapping it with L1 cache access\, but this overlap does not  \n hide the power dissipated by TLB lookups. It can even exacerbate the power  \n dissipation by requiring higher associativity L1 cache. With today's concern  \n for power dissipation\, designs could instead adopt a virtual L1 cache\,  \n wherein TLB access power is dissipated only after L1 cache misses.  \n Unfortunately\, virtual caches have compatibility issues\, such as supporting  \n writeable synonyms and x86’s physical page table walker. This work proposes  \n an Opportunistic Virtual Cache (OVC) that exposes virtual caching as a  \n dynamic optimization by allowing some memory blocks to be cached with virtual  \n addresses and others with physical addresses. OVC relies on small OS changes  \n to signal which pages can use virtual caching (e.g.\, no writeable synonyms)\,  \n but defaults to physical caching for compatibility. We show OVC's promise  \n with analysis that finds virtual cache problems exist\, but are dynamically  \n rare. We change 240 lines in Linux 2.6.28 to enable OVC. On experiments with  \n Parsec and commercial workloads\, the resulting system saves 94-99% of TLB  \n lookup energy and nearly 23% of L1 cache dynamic lookup energy.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9222.field_date.0.132
SUMMARY:Jai Menon: iGPU: Exception Support and Speculative Execution on GPUs
DTSTAMP:20130618T164412Z
DTSTART:20120606T210000Z
DTEND:20120606T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/igpu-exception-support-and-speculative-execution-gpus
LOCATION:1221 CS
DESCRIPTION:This is a practice talk for ISCA   Since the introduction of fully  \n programmable vertex shader hardware\, GPU computing has made tremendous  \n advances. Exception support and speculative execution are the next steps to  \n expand the scope and improve the usability of GPUs.  However\, traditional  \n mechanisms to support exceptions and speculative execution are highly  \n intrusive to GPU hardware design. This paper builds on two related insights  \n to provide a unified lightweight mechanism for supporting exceptions and  \n speculation on GPUs.   First\, we observe that GPU programs can be broken  \n into code regions that contain little or no live register state at their  \n entry point. We then also recognize that it is simple to generate these  \n regions in such a way that they are idempotent\, allowing their entry points  \n to function as program recovery points and enabling support for exception  \n handling\, fast context switches\, and speculation\, all with very low  \n overhead. We call the architecture of GPUs executing these idempotent  \n regions the \emph{iGPU} architecture. The hardware extensions required are  \n minimal and the  construction of idempotent code regions is fully  \n transparent under the typical dynamic compilation framework of GPUs.  We  \n demonstrate how iGPU exception support enables virtual memory paging with  \n very low overhead (1\% to 4\%)\, and how speculation support enables  \n circuit-speculation techniques that can provide over 25\% reduction in  \n energy.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9223.field_date.0.133
SUMMARY:Shravan Rayanchu: Models and Systems for Understanding Wireless Interference
DTSTAMP:20130618T164412Z
DTSTART:20120607T180000Z
DTEND:20120607T200000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/models-and-systems-understanding-wireless-interference
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: Suman Banerjee (advisor)\, Aditya Akella\, Paul Barford\, Ranveer  \n Chandra and Stark Draper
END:VEVENT
BEGIN:VEVENT
UID:calendar.9219.field_date.0.134
SUMMARY:Siddharth Barman: Approximation Algorithms fpr Network Design and  \n Partitioning Problems
DTSTAMP:20130618T164412Z
DTSTART:20120608T180000Z
DTEND:20120608T200000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/approximation-algorithms-fpr-network-design-and-partitioning-problems
LOCATION:4310 CS
DESCRIPTION:Committee:Shuchi Chawla (advisor)\, Eric Bach\, Stark Draper\, Benjamin Recht\,  \n Stephen Wright
END:VEVENT
BEGIN:VEVENT
UID:calendar.9241.field_date.0.135
SUMMARY:Kendrick Boyd: Unachievable Region in Precision-Recall Space and Its Effect  \n on Empirical Evaluation
DTSTAMP:20130618T164412Z
DTSTART:20120615T151500Z
DTEND:20120615T161500Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/unachievable-region-precision-recall-space-and-its-effect-empirical-evaluation
LOCATION:4765 MSC
DESCRIPTION:Practice talk for ICML 2012. Abstract: Precision-recall (PR) curves and the  \n areas under them are widely used to summarize machine learning results\,  \n especially for data sets exhibiting class skew. They are often used  \n analogously to ROC curves and the area under ROC curves. It is already known  \n that PR curves vary as class skew varies. What was not recognized before this  \n paper is that there is a region of PR space that is completely unachievable\,  \n and the size of this region varies only with the skew. This paper precisely  \n characterizes the size of that region and discusses its implications for  \n empirical evaluation methodology in machine learning.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9331.field_date.0.136
SUMMARY:Feng Niu: Web-scale Knowledge-base Construction via Statistical Inference and  \n Learning
DTSTAMP:20130618T164412Z
DTSTART:20120620T190000Z
DTEND:20120620T210000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/web-scale-knowledge-base-construction-statistical-inference-and-learning
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: Chris Re (advisor)\; AnHai Doan (advisor)\, Jude Shavlik
END:VEVENT
BEGIN:VEVENT
UID:calendar.9336.field_date.0.137
SUMMARY:Chong Sun: Multi-filter String Matching and Human-centric Entity Matching for  \n Information Extraction
DTSTAMP:20130618T164412Z
DTSTART:20120626T150000Z
DTEND:20120626T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/multi-filter-string-matching-and-human-centric-entity-matching-information-extraction
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: Jeffrey Naughton (advisor)\; Jignesh Patel\, Christopher Re\, Anhai  \n Doan\, Chad Navis
END:VEVENT
BEGIN:VEVENT
UID:calendar.9337.field_date.0.138
SUMMARY:Theophilus Benson: New Approaches to Managing Enterprise Networks
DTSTAMP:20130618T164412Z
DTSTART:20120629T180000Z
DTEND:20120629T200000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/new-approaches-managing-enterprise-networks
LOCATION:3310
DESCRIPTION:Committee: Aditya Akella (advisor)\; Suman Banerjee\, Michael Swift\, Jennifer  \n Rexford\, Parameswaran Ramanathan
END:VEVENT
BEGIN:VEVENT
UID:calendar.9232.field_date.0.139
SUMMARY:Leo Haller: Abstract Conflict Driven Clause Learning
DTSTAMP:20130618T164412Z
DTSTART:20120702T210000Z
DTEND:20120702T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/abstract-conflict-driven-clause-learning
LOCATION:4310 CS
DESCRIPTION:High performance propositional satisfiability solvers based on the Conflict  \n Driven Clause Learning framework (CDCL) have been a major driver of research  \n in areas such as verification and decision procedures. Modern satisfiability  \n procedures are routinely applied to problems that were deemed intractable  \n based on theoretical considerations 15 years ago. Lifting the algorithmic  \n lessons of CDCL to richer problem domains is a focus of ongoing research. In  \n this talk\, I show how one can generalise CDCL using lattice-theoretic  \n abstractions to yield natural domain SMT procedures for logics and program  \n verification problems. I leverage the simple insight that existing CDCL  \n solvers can be characterised as logical abstract interpreters. The resulting  \n Abstract CDCL (ACDCL) framework is a mathematical and algorithmic recipe for  \n lifting CDCL by combining over- and underapproximate abstractions. I discuss  \n the lattice-theoretic prerequisites of clause learning\, give conditions for  \n completeness and present two instantiations of ACDCL for program verification  \n and SMT problems which significantly outperform existing techniques. This  \n talk is based on joint work with Vijay D'Silva\, Daniel Kroening\, and Alberto  \n Griggio.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9374.field_date.0.140
SUMMARY:Willis Lang: Cost-Effective Cloud Data Processing
DTSTAMP:20130618T164412Z
DTSTART:20120709T143000Z
DTEND:20120709T163000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/cost-effective-cloud-data-processing
LOCATION:2310 CS
DESCRIPTION:Committee: Jignesh Patel (advisor)\;\,Jeffrey Naughton\, AnHai Doan\, Christopher  \n Re\, Jonathan Eckhardt
END:VEVENT
BEGIN:VEVENT
UID:calendar.9375.field_date.0.141
SUMMARY:Ting Chen: Indexing Text Documents for Fast Evaluation of Regular Expressions
DTSTAMP:20130618T164412Z
DTSTART:20120712T150000Z
DTEND:20120712T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/indexing-text-documents-fast-evaluation-regular-expressions
LOCATION:2310 CS
DESCRIPTION:Committee: AnHai Doan (advisor)\, Jin-Yi Cai\, Jeffrey Naughton\, Jignesh Patel\,  \n Christopher Re
END:VEVENT
BEGIN:VEVENT
UID:calendar.9379.field_date.0.142
SUMMARY:Derek Hower: Acoherent Shared Memory
DTSTAMP:20130618T164412Z
DTSTART:20120716T150000Z
DTEND:20120716T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/acoherent-shared-memory-0
LOCATION:3310
DESCRIPTION:Committee: Mark Hill (advisor)\, David Wood\, Michael Swift\, Benjamin Liblit\,  \n Mikko Lipasti
END:VEVENT
BEGIN:VEVENT
UID:calendar.9380.field_date.0.143
SUMMARY:Marc de Kruijf: Compiler Construction of Idempotent Regions and Applications  \n in Architechture Design
DTSTAMP:20130618T164412Z
DTSTART:20120720T150000Z
DTEND:20120720T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/compiler-construction-idempotent-regions-and-applications-architechture-design
LOCATION:2310 CS
DESCRIPTION:Committee: Kathikeyan Sankaralingam\, Mark Hill\, Gurindar Sohi\, Somesh Jha and  \n Mikko Lipasti
END:VEVENT
BEGIN:VEVENT
UID:calendar.9381.field_date.0.144
SUMMARY:Caroline Uhler: Geometry of the faithfulness assumption in causal inference
DTSTAMP:20130618T164412Z
DTSTART:20120725T210000Z
DTEND:20120725T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/geometry-faithfulness-assumption-causal-inference
LOCATION:Wisconsin Institute for...
DESCRIPTION:<?xml:namespace prefix = o />  Many algorithms for inferring causality rely  \n heavily on the faithfulness assumption. The main justification for imposing  \n the faithfulness assumption is that the set of unfaithful distributions has  \n Lebesgue measure zero\, since it can be seen as a collection of hypersurfaces.  \n However\, due to sampling error the faithfulness condition alone is not  \n sufficient for statistical estimation\, and strong-faithfulness has been  \n proposed and assumed to achieve uniform consistency. In contrast to the plain  \n faithfulness assumption\, the set of distributions that is not strong-faithful  \n has non-zero Lebesgue measure and in fact\, can be surprisingly large as we  \n show in this talk. We study the strong-faithfulness condition from the point  \n of view of real algebraic geometry and give upper and lower bounds on the  \n Lebesgue measure of strong-faithful distributions for various classes of  \n directed acyclic graphs. Our results imply fundamental limitations for  \n algorithms inferring causality based on partial correlations\, with the  \n PC-algorithm as its most prominent example.     
END:VEVENT
BEGIN:VEVENT
UID:calendar.9387.field_date.0.145
SUMMARY:Yeye He: Privacy Preserving Data Publishing and Analysis
DTSTAMP:20130618T164412Z
DTSTART:20120727T143000Z
DTEND:20120727T163000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/privacy-preserving-data-publishing-and-analysis
LOCATION:2310 computer Sciences
DESCRIPTION:Committee: Jeffrey Naughton (advisor)\, AnHai Doan\, Somesh Jha\, Rafael Lazimy\,  \n Christopher Re
END:VEVENT
BEGIN:VEVENT
UID:calendar.9392.field_date.0.146
SUMMARY:Houssam Nassif: Relational Differential Prediction
DTSTAMP:20130618T164412Z
DTSTART:20120803T190000Z
DTEND:20120803T210000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/relational-differential-prediction
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: David Page (advisor)\, Jude Shavlik\, Jerry Zhu\, Elizabeth Burnside\,  \n Vito Santos Costa
END:VEVENT
BEGIN:VEVENT
UID:calendar.9395.field_date.0.147
SUMMARY:Felix Herrmann: Compressive sensing and sparse recovery in exploration  \n seismology
DTSTAMP:20130618T164412Z
DTSTART:20120810T153000Z
DTEND:20120810T163000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/compressive-sensing-and-sparse-recovery-exploration-seismology
LOCATION:Wisconsin Institute for...
DESCRIPTION:Current-day imaging and full-waveform inversion technology increasingly  \n relies on faithful samplings and simulations of seismic wavefields. This  \n reliance on full sampling and high-fidelity wavefield simulations strains our  \n acquisition and processing systems and overcoming this impediment is becoming  \n one of the main challenges faced by our industry. By using randomized  \n dimensionality-reduction techniques\, we propose a new strategy where  \n acquisition and computational costs are no longer dictated by the sampling  \n grid but by transform-domain compressibility. To arrive at this result\, we  \n use recent insights from stochastic optimization\, statistical physics\, and  \n compressive sensing to reduce processing costs by minimizing the number of  \n required passes through the data by carrying out the inversions on random  \n subsets of data. By incorporating these ideas in the formulations of  \n reverse-time migration and full waveform inversion\, we are able to improve  \n the quality of their output at reduced computational costs.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9399.field_date.0.148
SUMMARY:Feng Niu: Web-scale Knowledge-base Construction via Statistical Inference and  \n Learning
DTSTAMP:20130618T164412Z
DTSTART:20120814T150000Z
DTEND:20120814T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/web-scale-knowledge-base-construction-statistical-inference-and-learning-0
LOCATION:4310
DESCRIPTION:Committee: Christopher Re (advisor)\, anHai Doan (advisor)\, Jude Shavlik\,  \n Jeffrey Naughton\, Mark Craven
END:VEVENT
BEGIN:VEVENT
UID:calendar.9402.field_date.0.149
SUMMARY:Venkatraman Govindaraju: Energy Efficient Computing through Compiler Assisted  \n Dynamic Specialization
DTSTAMP:20130618T164412Z
DTSTART:20120816T150000Z
DTEND:20120816T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/energy-efficient-computing-through-compiler-assisted-dynamic-specialization
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: Karthikeyan Sankaralingam (advisor)\, David Wood\, Mikko Lipasti\,  \n Somesh Jha\, Jignesh Patel
END:VEVENT
BEGIN:VEVENT
UID:calendar.9404.field_date.0.150
SUMMARY:Piramanayagam Arumuga Nainar: Applications of Static Analysis and Program  \n Structure in Statistical Debugging
DTSTAMP:20130618T164412Z
DTSTART:20120817T140000Z
DTEND:20120817T160000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/applications-static-analysis-and-program-structure-statistical-debugging
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: Benjamin Liblit (advisor) \, Susan Horwitz\, Shan Lu\, Thomas Reps\,  \n Xiaojin Zhu
END:VEVENT
BEGIN:VEVENT
UID:calendar.9412.field_date.0.151
SUMMARY:Cindy Rubio Gonzalez: Finding Error-Propagation Bugs in Complex Software  \n Using Static Analysis
DTSTAMP:20130618T164412Z
DTSTART:20120820T150000Z
DTEND:20120820T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/finding-error-propagation-bugs-complex-software-using-static-analysis
LOCATION:3310
DESCRIPTION:Committee: Benjamin Liblit (advisor) \, Remzi Arpaci-Dusseau\, Susan Horwitz\,  \n Shan Lu\, Thomas Reps       
END:VEVENT
BEGIN:VEVENT
UID:calendar.9405.field_date.0.152
SUMMARY:Jiexing LI: Progress Indicators for Database Queries
DTSTAMP:20130618T164412Z
DTSTART:20120820T190000Z
DTEND:20120820T210000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/progress-indicators-database-queries
LOCATION:2310 Computer Sciences
DESCRIPTION:Committee: Jeffrey Naughton (advisor)\; Jignesh Patel\, Christopher Re  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9413.field_date.0.153
SUMMARY:Aubrey Barnard: Identifying Causal Relationships in Structured Observational  \n Data
DTSTAMP:20130618T164412Z
DTSTART:20120823T150000Z
DTEND:20120823T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/identifying-causal-relationships-structured-observational-data
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: David Page (advisor)\, Jude Shavlik\, Christopher Re\, Vitor Santos  \n Costa                      
END:VEVENT
BEGIN:VEVENT
UID:calendar.9414.field_date.0.154
SUMMARY:New CS Grad Student Orientation
DTSTAMP:20130618T164412Z
DTSTART:20120827T140000Z
DTEND:20120827T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/new-cs-grad-student-orientation
LOCATION:1240 Computer Sciences
END:VEVENT
BEGIN:VEVENT
UID:calendar.9458.field_date.0.155
SUMMARY:Barış Aydınlıoğlu: A Study of Computational Hardness and Randomness
DTSTAMP:20130618T164412Z
DTSTART:20120829T190000Z
DTEND:20120829T210000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/study-computational-hardness-and-randomness
LOCATION:CS 4310
DESCRIPTION:Committee: Eric Bach\, Jin-Yi Cai\, Dieter van Melkebeek (advisor)
END:VEVENT
BEGIN:VEVENT
UID:calendar.9517.field_date.0.156
SUMMARY:Chaitanya Gokhale: Crowdsourced Entity Matching
DTSTAMP:20130618T164412Z
DTSTART:20120829T190000Z
DTEND:20120829T210000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/crowdsourced-entity-matching
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: AnHai Doan (advisor)\, Jeff Naughton\, Jerry Zhu
END:VEVENT
BEGIN:VEVENT
UID:calendar.10357.field_date.0.157
SUMMARY:Matthew Fredrikson: Reasoning About Information Leakage and Adversarial  \n Inference
DTSTAMP:20130618T164412Z
DTSTART:20120905T140000Z
DTEND:20120905T160000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/reasoning-about-information-leakage-and-adversarial-inference
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: Somesh Jha (advisor)\; Thomas Reps\, Thomas Ristenpart  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9401.field_date.0.158
SUMMARY:James (Jim) Luedtke\, Assistant Professor: Branch-and-cut approaches for  \n chance-constrained formulations of reliable network design problems
DTSTAMP:20130618T164412Z
DTSTART:20120910T210000Z
DTEND:20120910T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/branch-and-cut-approaches-chance-constrained-formulations-reliable-network-design-problems
LOCATION:Wisconsin Institute for...
DESCRIPTION:We study the design of reliably connected networks. Given a graph with arcs  \n that may fail at random\, the goal is to select a minimum cost set of arcs  \n such that a connectivity requirement is met with high probability. We first  \n compare this model with a well-known deterministic model of reliable network  \n design: survivable network design. We demonstrate that\, if distributional  \n information on arc failures is known\, the chance constraint model can yield a  \n significantly richer set of solutions on the efficient frontier of  \n reliability and cost. We then present two solution approaches for our model\,  \n which we formulate as a chance-constrained stochastic integer program. The  \n first approach is based on a formulation that uses binary variables to  \n determine if the connectivity requirement is satisfied in each arc failure  \n scenario\, and enforces the connectivity requirement in selected scenarios  \n using scenario-based graph cuts. We derive additional classes of valid  \n inequalities for this formulation and study their facet-inducing properties.  \n The second formulation is based on the idea of probabilistic graph cuts\,  \n which is an extension of graph cuts to graphs with random arc failures.  \n Inequalities defined by such cuts are sufficient to define the set of  \n feasible solutions and can be separated efficiently at integer solutions\,  \n allowing this formulation to be solved by a branch-and-cut algorithm.  \n Computational results will be presented that demonstrate that the approaches  \n can solve large instances. This is joint work with Yongjia Song.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9519.field_date.0.159
SUMMARY:Jin-Yi Cai: Madison Chaos and Complex Systems Seminar
DTSTAMP:20130618T164412Z
DTSTART:20120911T170500Z
DTEND:20120911T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/madison-chaos-and-complex-systems-seminar
LOCATION:4274 Chamberlin Hall
DESCRIPTION:http://sprott.physics.wisc.edu/chaos-complexity/12fall.html   September 11\,  \n 2012 Computational complexity theory --- The world of P and NP Jin-Yi Cai\, UW  \n Department of Computer Sciences Computational Complexity Theory is the study  \n of intrinsic difficulties of computational problems. The most prominent open  \n problem is the conjecture that P is not equal to NP.  In essense this  \n conjecture states that it is intrinsically harder to find proofs than to  \n verify them. It has a fundamental importance in many areas from computer  \n science to mathematics\, to our basic understanding of nature. Valiant's new  \n theory of holographic algorithms is one of the most beautiful ideas in  \n algorithm design in recent memory.  It gives a new look on the P versus NP  \n problem. In this theory\, information is represented by a superposition of  \n linear vectors in a holographic mix. This mixture creates the possibility for  \n exponential sized cancellations of fragments of local computations. The  \n underlying computation is done by invoking the Fisher-Kasteleyn-Temperley  \n method for counting perfect matchings for planar graphs (Dimer Problem).  \n Holographic algorithms challenge our conception of what polynomial time  \n computation can do\, in view of the P vs. NP question. In this talk we will  \n survey the developments in holographic algorithms. No specialized background  \n is assumed.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10420.field_date.0.160
SUMMARY:New Grad Student Social
DTSTAMP:20130618T164412Z
DTSTART:20120911T213000Z
DTEND:20120911T233000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/new-grad-student-social
LOCATION:Union Sputh Varsity Hall...
END:VEVENT
BEGIN:VEVENT
UID:calendar.10360.field_date.0.161
SUMMARY:Ajay Gulati: Distributed Resource Scheduler (DRS): Design\, Implementation\,  \n and Lessons Learned
DTSTAMP:20130618T164412Z
DTSTART:20120913T170000Z
DTEND:20120913T183000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/distributed-resource-scheduler-drs-design-implementation-and-lessons-learned
LOCATION:Union South - 5th...
DESCRIPTION:NOTE: Lunch will be served at noon\; the talk will start some time thereafter  \n (perhaps 12:20 or so). Abstract: Automated management of physical resources  \n is critical for reducing the operational costs of virtualized environments.  \n An effective resource-management solution must provide performance isolation  \n among virtual machines (VMs)\, handle resource fragmentation across physical  \n hosts and optimize scheduling for multiple resources. It must also utilize  \n the underlying hardware infrastructure efficiently. In this talk\, I will  \n present the design and implementation of one of our key management solution:  \n DRS. I will also highlight some key lessons learned from production customer  \n deployments over a period of more than five years. VMware’s Distributed  \n Resource Scheduler (DRS) manages the allocation of physical resources to a  \n set of virtual machines deployed in a cluster of hosts\, each running the  \n VMware ESX hypervisor. DRS maps VMs to hosts and performs intelligent load  \n balancing in order to improve performance and to enforce both user-specified  \n policies and system-level constraints. DRS also supports a “what-if”  \n mode\, making it possible to evaluate the impact of changes in workloads or  \n cluster configuration. Bio: Ajay Gulati is an R & D staff member in the  \n distributed resource management team at VMware. At VMware\, his work has lead  \n to some of the new storage management features such as Storage I/O control  \n and Storage DRS\, released as part of vSphere 4.1 and 5.0 releases  \n respectively. Along with developing products at VMware\, he has published his  \n research at many conferences such as SOCC\, OSDI\, Sigmetrics\, Usenix ATC\,  \n Usenix FAST and SPAA. He has also given several talks on various  \n storage-related topics at VMworld\, which is an industry conference on  \n virtualization.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9524.field_date.0.162
SUMMARY:Greg Wright: Mobile = low power? A server guy's journey to the low end
DTSTAMP:20130618T164412Z
DTSTART:20120913T210000Z
DTEND:20120913T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/mobile-low-power-server-guys-journey-low-end
LOCATION:1221 CS
DESCRIPTION:Are mobile systems just like low power servers?  And how much power  \n is "high" for a smartphone anyway?  This talk explores the  \n application characteristics vs. energy and power constraints on mobile  \n computing devices\, with some architectural implications and opportunities  \n for research.   Bio: Greg Wright is a Senior Staff Engineer/Manager in  \n Qualcomm Research\, currently moving from Silicon Valley to a new  \n processor-focused research location in Raleigh\, NC.  His research interests  \n include processor architecture and virtual machines.  Prior to joining  \n Qualcomm in 2010\, Greg worked almost 10 years in Sun Labs on  \n processor/memory system.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9378.field_date.0.163
SUMMARY:Rong Jin\, Associate Professor: A Simple Algorithm for Semi-supervised  \n Learning with Improved Generalization Error Bound
DTSTAMP:20130618T164412Z
DTSTART:20120914T193000Z
DTEND:20120914T203000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/simple-algorithm-semi-supervised-learning-improved-generalization-error-bound
LOCATION:CS4310
DESCRIPTION:In this talk\, I will present a simple algorithm for semi-supervised  \n regression. The key idea is to use the top eigenfunctions of integral  \n operator derived from unlabeled examples as the basis functions and learn the  \n prediction function by a simple linear regression. We show that under  \n appropriate assumptions about the integral operator\, this approach is able to  \n achieve a regression error that is close to the optimal. We also verify the  \n effectiveness of the proposed algorithm by an empirical study. Bio: Rong Jin  \n focuses his research on statistical machine learning and its application to  \n information retrieval. He has worked on a variety of machine learning  \n algorithms and their application to information retrieval\, including  \n retrieval models\, collaborative filtering\, cross lingual information  \n retrieval\, document clustering\, and video/image retrieval. He has published  \n over 160 conference and journal articles on related topics. He is currently  \n an associate editor of ACM Transaction on Data Mining and Knowledge Discovery  \n (KDD). Dr. Jin holds a Ph.D. in Computer Science from Carnegie Mellon  \n University He received the NSF Career Award in 2006.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10359.field_date.0.164
SUMMARY:Junming Sui: Socioscope: Spatio-Temporal Signal Recovery from Social Media
DTSTAMP:20130618T164412Z
DTSTART:20120917T210000Z
DTEND:20120917T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/socioscope-spatio-temporal-signal-recovery-social-media
LOCATION:CS 4310
DESCRIPTION:This is a practical talk for our ECML-PKDD 2012 paper:   Socioscope:  \n Spatio-Temporal Signal Recovery from Social Media   Jun-Ming Xu\, Aniruddha  \n Bhargava\, Robert Nowak and Xiaojin Zhu   Abstract: Many real-world phenomena  \n can be represented by a spatio-temporal signal: where\, when\, and how  \n much. Social media is a tantalizing data source for those who wish to  \n monitor such signals. Unlike most prior work\, we assume that the target  \n phenomenon is known and we are given a method to count its occurrences in  \n social media. However\, counting is plagued by sample bias\, incomplete data\,  \n and\, paradoxically\, data scarcity -- issues inadequately addressed by prior  \n work. We formulate signal recovery as a Poisson point process estimation  \n problem. We explicitly incorporate human population bias\, time delays and  \n spatial distortions\, and spatio-temporal regularization into the model to  \n address the noisy count issues. We present an efficient optimization  \n algorithm and discuss its theoretical properties. We show that our model is  \n more accurate than commonly-used baselines. Finally\, we present a case study  \n on wildlife roadkill monitoring\, where our model produces qualitatively  \n convincing results.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.10358.field_date.0.165
SUMMARY:John C. Doyle: J. Barkley Rosser Memorial Lecture: Universal Laws and  \n Architectures
DTSTAMP:20130618T164412Z
DTSTART:20120918T210000Z
DTEND:20120918T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/j-barkley-rosser-memorial-lecture-universal-laws-and-architectures
LOCATION:Forum (1st Floor) at the...
DESCRIPTION:A reception will follow the talk in the WID Atrium Abstract: This talk will  \n focus on progress towards a more “unified” theory for complex networks  \n motivated by biology and technology. This unification involves several  \n elements: hard limits on achievable robust performance (“laws”)\, the  \n organizing principles that succeed or fail in achieving them (architectures  \n and protocols)\, the resulting high variability data and “robust yet  \n fragile” behavior observed in real systems (behavior\, data)\, and the  \n processes by which systems evolve (variation\, selection\, design). We will  \n leverage a series of case studies from neuroscience\, cell biology\, human  \n physiology\, and technology to illustrate the implications of recent  \n theoretical developments. Hard limits on measurement\, prediction\,  \n communication\, computation\, decision\, and control\, as well as the underlying  \n physical energy and material conversion mechanism necessary to implement  \n these abstract process are at the heart of modern mathematical theories of  \n systems in engineering and science (often associated with names such as  \n Shannon\, Poincare\, Turing\, Gödel\, Bode\, Wiener\, Heisenberg\, Carnot\, et  \n cetera). They form the foundation for rich and deep subjects that are  \n nevertheless now introduced at the undergraduate level. Unfortunately\, these  \n subjects remain largely fragmented and incompatible\, even as the tradeoffs  \n between these limits are essential to understanding human physiology and  \n neuroscience\, and are of growing importance in building integrated and  \n sustainable systems. Time permitting\, we will give an accessible introduction  \n to these theories\, how they do and don’t relate to each other\, and progress  \n and prospects for a more integrated theory. Particular emphasis will be put  \n on Turing’s work in honor of his 100th birthday. Bio: John C. Doyle is the  \n John G Braun Professor of Control and Dynamical Systems\, Electrical  \n Engineering\, and BioEngineering at Caltech. He has a BS and MS in EE from MIT  \n (1977)\, and a PhD in Math from UC Berkeley (1984). Current research interests  \n are the theoretical foundations for complex networks in engineering and  \n biology and for multiscale physics. Early work was in the mathematics of  \n robust control\, including extensions to nonlinear and networked systems.  \n Related software projects include the Robust Control Toolbox (muTools)\,  \n SOSTOOLS\, SBML (Systems Biology Markup Language)\, and FAST (Fast AQM\,  \n Scalable TCP). Prize papers include IEEE Baker\, IEEE Automatic Control  \n Transactions Axelby (twice)\, and best conference papers in ACM Sigcomm and  \n AACC American Control Conference. Individual awards include AACC Eckman\, and  \n IEEE Control Systems Field and Centennial Outstanding Young Engineer Awards.  \n He has held national and world records and championships in various sports.  \n He is best known for having excellent co-authors\, students\, friends\, and  \n colleagues.
END:VEVENT
BEGIN:VEVENT
UID:calendar.9454.field_date.0.166
SUMMARY:Hayri Önal\, Professor : Mathematically Correct Political Districting
DTSTAMP:20130618T164412Z
DTSTART:20120924T210000Z
DTEND:20120924T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/mathematically-correct-political-districting
LOCATION:Wisconsin Institute for...
DESCRIPTION:Political districting problem is one of the most difficult combinatorial  \n optimization problems. It involves various spatial and social considerations\,  \n including population equality\, spatial contiguity\, compactness\, and community  \n integrity\, which typically conflict with each other. In this talk a large  \n scale linear integer programming model is presented where these four criteria  \n are incorporated simultaneously to draw the boundaries of state legislative  \n districts in Illinois using the 2000 and 2010 Census data. The results show  \n that the model-generated districts are considerably more compact than the  \n actual districts in the State’s districting plan implemented in the  \n elections since 2001 while simultaneously dividing counties much less and  \n improving minority representation.   Biography: Hayri Önal received his BS  \n and MS in Mathematics and PhD in Operations Research\, all from the Middle  \n East Technical University (METU)\, Ankara\, Turkey. He served as a faculty  \n member in the Department of Mathematics and Department of Industrial  \n Engineering of METU before joining the University of Illinois at  \n Urbana-Champaign. Currently he is a professor in the Department of  \n Agricultural and Consumer Economics. He teaches mathematics for economists\,  \n applied mathematical programming\, and dynamic simulation courses. His recent  \n research focuses on spatial optimization and environmental and resource  \n economics with emphasis on conservation reserve design and renewable energy.  \n  
END:VEVENT
BEGIN:VEVENT
UID:calendar.10450.field_date.0.167
SUMMARY:Bryan Gibson: Using Machine Learning to Understand and Influence Human  \n Categorization Behavior
DTSTAMP:20130618T164412Z
DTSTART:20120925T190000Z
DTEND:20120925T210000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/using-machine-learning-understand-and-influence-human-categorization-behavior
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: Jerry Zhu (advisor\, Mark Craven\, Bilge Mutlu\, Tim Rogers\, Benjamin  \n Snyder
END:VEVENT
BEGIN:VEVENT
UID:calendar.10459.field_date.0.168
SUMMARY:Christopher Rao: Design and Diversity in Bacterial Signaling Networks
DTSTAMP:20130618T164412Z
DTSTART:20120925T210000Z
DTEND:20120925T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/design-and-diversity-bacterial-signaling-networks
LOCATION:1227 Engineering Hall
DESCRIPTION:Computation & Informatics in Biology & Medicine (CIBM) Seminar Abstract:  \n Complex arrays of interconnected pathways govern the behavior of many  \n cellular processes\, including those used to survive in hostile environments  \n and to invade other organisms. How cells decide which processes to activate  \n and which ones to deactivate is most often still unknown. What is known is  \n that these decisions are often mediated by interlocking positive and negative  \n feedback loops. These feedback loops are thought to dynamically coordinate  \n complex cellular responses to different environmental signals. In this talk\,  \n I will discuss recent work where we have characterized the role of  \n interlocking feedback loops in regulating a number of cellular processes.  \n Starting with two simple systems involved in the regulation of antibiotic  \n resistance in bacteria\, I will show how interlocking feedback loops are used  \n to shape the response and governing regulation of these systems. I will then  \n extend these ideas to two more complicated systems involved in bacterial  \n pathogenesis and chemotaxis.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10470.field_date.0.169
SUMMARY:Karu Sankaralingam: Brave new world of Arduino and Pedagogical Uses
DTSTAMP:20130618T164412Z
DTSTART:20120925T210000Z
DTEND:20120925T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/brave-new-world-arduino-and-pedagogical-uses
LOCATION:1240 CS
DESCRIPTION:In this talk I will describe some of my experiences in using the Ardiuno  \n board and its ecosystem of controllers and shields which we used to create  \n handson projects in the Freshman computer science class. I will describe  \n some of the student feedback\, learning objectives\, pedagogical lessons I  \n learned\, and my thoughts on how such activities could enhance the  \n CS curriculum.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10488.field_date.0.170
SUMMARY:Mohammed Hidayath Ansari: Using In Situ Statistics to track White Matter  \n Changes in the Brain
DTSTAMP:20130618T164412Z
DTSTART:20120927T150000Z
DTEND:20120927T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/using-situ-statistics-track-white-matter-changes-brain
LOCATION:3310 CS
DESCRIPTION:Committee: Michael Coen (advisor)\, David Page\, Barbara Bendlin
END:VEVENT
BEGIN:VEVENT
UID:calendar.9441.field_date.0.171
SUMMARY:Alex Wyler: Hacking the Graph
DTSTAMP:20130618T164412Z
DTSTART:20120927T210000Z
DTEND:20120927T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/hacking-graph
LOCATION:1240CS
DESCRIPTION:At Facebook's core is the social graph: people and the connections they have  \n to everything they care about. Historically\, Facebook has managed this graph  \n and has expanded it over time as we launch new products (ex: photos\, places).  \n  The graph has now been extended to include arbitrary actions and objects  \n created by third-party apps\, enabling these apps to integrate deeply into the  \n Facebook experience.  This talk will focus on the motivations behind and  \n technical details surrounding Facebook's Open Graph product.   Alex attended  \n UW-Madison as CS and ECE double major\, starting in fall of 2007. He  \n interviewed with Facebook after one of the "hackathons" on campus\, and  \n accepted a summer internship position.  After leaving school\, he returned to  \n Facebook as a fulltime software engineer.  As a student\, he co-founded   \n http://madbites.com  (formerly badgerbites.com)\, a local online  \n food-ordering technology.  The company (now officially named  UConnect   \n http://uconnectfood.com/corporate/ ) is going strong and has expanded to 13  \n cities throughout the country.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10469.field_date.0.172
SUMMARY:Daniel Luchaup: Programming Languages and Probability Techniques with  \n Applications in Android Security
DTSTAMP:20130618T164412Z
DTSTART:20120928T144500Z
DTEND:20120928T164500Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/programming-languages-and-probability-techniques-applications-android-security
LOCATION:2310
DESCRIPTION:Committee:Somesh Jha (advisor)\, Thomas Reps\, Thomas Ristenpart
END:VEVENT
BEGIN:VEVENT
UID:calendar.10471.field_date.0.173
SUMMARY:Aditya Akella\, Andrea Arpaci-Dusseau\, Remzi Arpaci-Dusseau\, Paul Barford\,  \n Somesh Jha\, Tom Ristenpart\, Mike Swift: WISDoM: Introduction and Overview
DTSTAMP:20130618T164412Z
DTSTART:20120928T210000Z
DTEND:20120928T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/wisdom-introduction-and-overview
LOCATION:1221
DESCRIPTION:The future infrastructure of the data center is increasingly becoming  \n "software defined": instead of adhering to (or even being aware of) the  \n limits of the underlying physical infrastructure\, applications and platforms  \n will be able to specify their fine-grained needs\, thus precisely defining the  \n virtual environment in which they wish to run\; software\, placed at key  \n locations within the datacenter infrastructure\, will implement the needed  \n functionality\, thus delivering hardware-like features to applications and  \n services in the manner they desire\; the various components that make up such  \n a system will be deeply programmable\, facilitating the realization of such  \n systems. This "software-defined datacenter" (SDD) thus forms the core of the  \n future cloud platform. To tackle these problems\, and usher in a new era of  \n the software-defined data center\, we are forming the Wisconsin Institute on  \n Software-defined Datacenters in Madison (WISDoM). The institute brings  \n together an eclectic group of researchers in the Computer Sciences department  \n in storage\, networks\, and security. The storage team consists of Professors  \n Andrea C. Arpaci-Dusseau\, Remzi H. Arpaci-Dusseau\, and Michael Swift\;  \n Professors Aditya Akella and Paul Barford form the networking team\;  \n Professors Somesh Jha and Thomas Ristenpart comprise the security team. This  \n "micro-interdisciplinary" group\, reaching across three distinct areas of  \n computer science\, has been brought together with the single focus of defining  \n the future SDD and attacking the research challenges therein\; because the  \n problems within SDDs are large and complex\, we believe such a broad team is  \n critical to success. WISDoM is a targeted effort to tackle the research  \n problems within the SDD environment over the next four to five years. In this  \n time\, we will address the key challenges in SDDs\, train and educate large  \n numbers of graduate and undergraduate students in the relevant topics\, and  \n help further computer science in general through outreach activities. These  \n three missions (research\, education\, and outreach) form the core of the  \n institute.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10356.field_date.0.174
SUMMARY:Rebecca Fiebrink\, Assistant Professor of Computer Science (also Music)\, :  \n Interactive Machine Learning in Music Performance and Composition
DTSTAMP:20130618T164412Z
DTSTART:20121001T210000Z
DTEND:20121001T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/interactive-machine-learning-music-performance-and-composition
LOCATION:CS Room 1240
DESCRIPTION:In my work at Princeton\, I build\, study\, teach about\, and perform with new  \n human-computer interfaces for real-time digital music. Much of my work has  \n concerned the use of supervised learning as a tool for musicians\, artists\,  \n and composers to build digital musical instruments and other real-time  \n interactive systems. Through the use of training data\, supervised learning  \n algorithms offer users a means to specify the relationship between low-level\,  \n human-generated control signals (such as the outputs of  \n gesturally-manipulated sensor interfaces\, or audio captured by a microphone)  \n and the desired computer response (such as a change in the parameters that  \n dynamically drive computer-generated audio). The task of creating an  \n interactive system can therefore be formulated not as a task of writing and  \n debugging code\, but rather one of designing and revising a set of training  \n examples that implicitly encode a target function\, and of choosing and tuning  \n an algorithm to learn that function. In this talk\, I will provide a brief  \n introduction to interactive computer music and the use of supervised learning  \n in this field. I will show a live musical demo of the software that I have  \n created to enable non-computer-scientists to interactively apply standard  \n supervised learning algorithms to music and other real-time problem domains.  \n This software\, called the Wekinator\, supports human interaction throughout  \n the entire supervised learning process\, including the generation of training  \n data by real-time demonstration and the evaluation of trained models through  \n hands-on application to real-time inputs. Drawing on my work with users  \n applying the Wekinator to real-world problems\, I'll discuss how data-driven  \n methods can enable more effective approaches to building interactive systems\,  \n through supporting rapid prototyping and an embodied approach to design\, and  \n through "training" users to become better machine learning practitioners.  \n I'll also discuss some of the remaining challenges at the intersection of  \n machine learning and human-computer interaction that must be addressed for  \n end users to apply machine learning more efficiently and effectively\,  \n especially in interactive contexts.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10490.field_date.0.175
SUMMARY:Matthew Renzelmann: Systems/Security Seminar
DTSTAMP:20130618T164412Z
DTSTART:20121001T210000Z
DTEND:20121001T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/systemssecurity-seminar
LOCATION:2310
DESCRIPTION:Matt Renzelmann will do a practice talk for his upcoming OSDI paper\,  \n "SymDrive:  Testing Drivers without Devices."  This paper is joint work  \n with Asim Kadav and Mike Swift here at UW.  The talk is 25 minutes with time  \n for questions\, feedback\, and discussion afterward.  Read more at  \n http://research.cs.wisc.edu/sonar/projects/symdrive/ Abstract: Device-driver  \n development and testing is a complex and error-prone undertaking. For  \n example\, testing error handling code requires simulating faulty inputs from  \n the device. A single driver may support dozens of devices\, and a developer  \n may not have access to any of them. Consequently\, many Linux driver patches  \n include the comment "compile tested only." SymDrive is a system for testing  \n Linux and FreeBSD drivers without their devices present. The system uses  \n symbolic execution to remove the need for hardware\, and extends past tools  \n with three new features. First\, SymDrive uses static analysis and  \n source-to-source transformation to greatly reduce the effort of testing a new  \n driver. Second\, SymDrive checkers are ordinary C code and execute in the  \n kernel\, where they have full access to kernel and driver state. Finally\,  \n SymDrive provides an execution tracing tool to identify how a patch changes  \n I/O to the device and to compare device-driver implementations. In applying  \n SymDrive to 21 Linux drivers and 5 FreeBSD drivers\, we found 39 bugs.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10495.field_date.0.176
SUMMARY:Guoliang Jin: Automated Concurrency-Bug Fixing
DTSTAMP:20130618T164412Z
DTSTART:20121002T200000Z
DTEND:20121002T210000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/automated-concurrency-bug-fixing
LOCATION:3310
DESCRIPTION:  Guoliang Jin will do a practice talk for his upcoming OSDI paper\,  \n "Automated Concurrency-Bug Fixing". The talk will be 25 minutes with time for  \n questions\, feedback\, and discussion afterward.   Abstract:   Concurrency  \n bugs are widespread in multithreaded programs. Fixing them is time-consuming  \n and error-prone. We present CFix\, a system that automates the repair of  \n concurrency bugs. CFix works with a wide variety of concurrency-bug  \n detectors. For each failure-inducing interleaving reported by a bug detector\,  \n CFix first determines a combination of mutual-exclusion and order  \n relationships that\, once enforced\, can prevent the buggy interleaving. CFix  \n then uses static analysis and testing to determine where to insert what  \n synchronization operations to force the desired mutual-exclusion and order  \n relationships\, with a best effort to avoid deadlocks and excessive  \n performance losses. CFix also simplifies its own patches by merging fixes for  \n related bugs.   Evaluation using four different types of bug detectors and  \n thirteen real-world concurrency-bug cases shows that CFix can successfully  \n patch these cases without causing deadlocks or excessive performance  \n degradation. Patches automatically generated by CFix are of similar quality  \n to those manually written by developers.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.9525.field_date.0.177
SUMMARY:Boris Grot: Scale out Processors
DTSTAMP:20130618T164412Z
DTSTART:20121002T210000Z
DTEND:20121002T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/scale-out-processors
LOCATION:1240 CS
DESCRIPTION:A growing number of today’s most relevant applications are served online  \n and run in large-scale datacenters characterized by thousands of servers and  \n multi-megawatt power budgets. As Dennard scaling comes to a halt\, experts are  \n projecting exponential growth in datacenter power and performance  \n requirements in the coming decade\, driven by the rising popularity of the  \n online service model. To efficiently meet the computing needs in the  \n post-Dennard era\, datacenters will rely on a new form of ISA – Integration\,  \n Specialization\, and Approximation. As a first step toward this post-Dennard  \n ISA\, we have developed Scale-Out Processors – a processor design  \n methodology that maximizes performance per TCO on scale-out workloads running  \n in large-scale datacenters. Using a metric of performance density\, our  \n methodology facilitates the design of optimal configurations\, called pods\, of  \n cores\, caches\, and interconnect. Each pod is a stand-alone server-on-chip\, a  \n feature that avoids the expense and complexity of global (i.e.\, inter-pod)  \n interconnect and coherence. As I will demonstrate\, Scale-Out Processors yield  \n higher performance\, lower TCO\, and better technology scalability over  \n existing design alternatives. Bio: Boris Grot is a post-doctoral researcher  \n in the Parallel Systems Architecture Lab at EPFL. His research focuses on  \n improving the efficiency of large-scale datacenters through advancements to  \n server processor architectures\, memory systems\, and interconnects. Grot  \n received his PhD in Computer Science from The University of Texas at Austin  \n in 2011. His thesis addressed challenges of scalability and  \n quality-of-service in on-chip networks of highly-integrated processor chips.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10494.field_date.0.178
SUMMARY:Benjamin Mueller: Modeling Cα–H∙∙∙O Mediated Transmembrane Dimers
DTSTAMP:20130618T164412Z
DTSTART:20121002T210000Z
DTEND:20121002T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/modeling-c%CE%B1%E2%80%93h%E2%88%99%E2%88%99%E2%88%99o-mediated-transmembrane-dimers
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) Seminar Abstract:  \n Computational structural prediction can assist in the study of proteins in  \n the absence of experimentally determined structural data. This work focuses  \n on an ab initio method that can effectively predict the structure of  \n interacting transmembrane helices.  The method applies to transmembrane  \n associations mediated by networks of weak hydrogen bonds that form between  \n Cα donors on one helix and backbone carbonyl groups on the other helix  \n (Cα–H∙∙∙O=C hydrogen bonds). Specific backbone arrangements are  \n required to orient the helices so that the geometry is compatible with  \n interhelical hydrogen bond formation.  Due to these restrictions the  \n structural motifs are more recognizable and provide a number of constraints  \n making their prediction at near atomic level amenable. Our method is based on  \n a pre-computed analysis of the conformational space that permits the  \n formation of Cα–H∙∙∙O=C-mediated interhelical interactions\, which  \n substantially limits the search space. The result is a high-throughput ab  \n initio method that can predict the structure of known transmembrane  \n homodimers with near atomic precision. The rapidity of the method allows us  \n to perform genome wide searches for Cα–H∙∙∙O=C-mediated interhelical  \n dimers. The method provides an opportunity for building\, validating and  \n analyzing a comprehensive atlas of transmembrane interactions for entire  \n genomes.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.10382.field_date.0.179
SUMMARY:Rajiv Batra\, Pavan Nigam\, Ramu Sunkara: Entrepreneurship Experiences\, the  \n Badger Entrepreneurship Forum\, and Internship Opportunities for Students
DTSTAMP:20130618T164412Z
DTSTART:20121004T210000Z
DTEND:20121004T223000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/entrepreneurship-experiences-badger-entrepreneurship-forum-and-internship-opportunities-studen
LOCATION:1240
DESCRIPTION:This presentation will have three parts. First\, UW-CS Alum and Palo Alto  \n Networks co-founder and VP of Engineering Rajiv Batra will give a talk "My  \n Entrepreneurial Journey."  In it he will talk about his path from UW-student  \n all the way to founding and building Palo Alto Networks. Palo Alto Networks  \n had one of the most successful IPOs in 2012 (www.paloaltonetworks.com). Next\,  \n UW-CS Alum Pavan Nigam will talk about a group he founded in the San  \n Francisco Bay Area\, the "Badger Entrepreneurship Forum"  \n  (http://www.siliconvalleybadgers.com/)\, a community of UW-Alums that  \n includes entrepreneurs\, CEOs\, executives\, VCs\, investment bankers\, attorneys  \n and other service providers to the entrepreneurial ecosystem. Pavan was  \n co-founder of Healtheon\, which is now WebMD. Finally\, UW-CS Alum Ramu Sunkara  \n will talk about an exciting opportunity for student internships in startup  \n companies in the bay area that he is developing in conjunction with the CS  \n department. Ramu was the founder of Qik (http://qik.com/)\, which was acquired  \n by Skype in 2011.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10503.field_date.0.180
SUMMARY:Paul Black: tba
DTSTAMP:20130618T164412Z
DTSTART:20121004T210000Z
DTEND:20121004T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/tba-0
LOCATION:1240CS
DESCRIPTION:tba
END:VEVENT
BEGIN:VEVENT
UID:calendar.10474.field_date.0.181
SUMMARY:Spiro Maroulis\, Assistant Professor\, School of Public Affairs & Associate  \n Director of Policy Informatics\, ASU Decision Theater: Social and Task  \n Interdependencies in the Frontline Implementation of Innovation
DTSTAMP:20130618T164412Z
DTSTART:20121008T210000Z
DTEND:20121008T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/social-and-task-interdependencies-frontline-implementation-innovation
LOCATION:Wisconsin Institute for...
DESCRIPTION:Organizations often adopt\, though not successfully implement\, innovations  \n that require widespread frontline participation to realize their full  \n benefits. To better understand the mechanisms that support or inhibit the  \n implementation of such innovations\, we apply constructs from organizational  \n learning theory to the case of implementing innovation in schools\, and  \n operationalize those constructs in a computational\, agent-based model. Model  \n analysis reveals that individual-level improvement before the decision to  \n adopt an innovation can lead to an unanticipated side effect – the  \n elimination of work activities needed to support the implementation phase of  \n the adoption process. We demonstrate how this side effect can impede  \n frontline implementation even when the individuals inside the organization  \n are able and willing to locally implement the constituent tasks of an  \n innovation. We also translate our insights into specific propositions about  \n how the work required by an innovation and an organization’s internal  \n social network relate to the level of frontline implementation in  \n organizations\, and suggest managerial leverage points for improving the  \n execution of intentional change efforts.   \n http://www.public.asu.edu/~smarouli/
END:VEVENT
BEGIN:VEVENT
UID:calendar.10508.field_date.0.182
SUMMARY:Kendrick Boyd: Precision-Recall Space Properties and Applications for Machine  \n Learning
DTSTAMP:20130618T164412Z
DTSTART:20121009T180000Z
DTEND:20121009T200000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/precision-recall-space-properties-and-applications-machine-learning
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: David Page (advisor)\; Elizabeth Burnside\, Mark Craven\, Jude  \n Shavlik\, Jerry Zhu  
END:VEVENT
BEGIN:VEVENT
UID:calendar.10435.field_date.0.183
SUMMARY:David Hansquine: Computing Trends & The Mobile Experience
DTSTAMP:20130618T164412Z
DTSTART:20121009T210000Z
DTEND:20121009T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/computing-trends-mobile-experience
LOCATION:1240 CS
DESCRIPTION:    The processor powering your mobile phone may soon outperform the one  \n in your laptop or even desktop.  While years ago\, it was inconceivable that  \n GHz processors could be utilized let alone manufactured for such form  \n factors\, today they are becoming ubiquitous.  However\, 10+ GHz of combined  \n processor performance will not matter without also enabling a more  \n compelling mobile experience while addressing the challenges this creates.  \n  This talk will cover current trends and technologies seeking to utilize  \n the available processing power and possible solutions to address the  \n challenges imposed by the mobile form factor. Speaker Bio:  Since joining  \n Qualcomm in 1995\, David Hansquine has designed various wireless modem and  \n microprocessor-related blocks while leading more than a dozen ASICs for both  \n mobile handsets and infrastructure products.  He was responsible for  \n several of Qualcomm’s first WCDMA chips\, then later led the team  \n delivering Qualcomm’s first GHz and quad-core application processors for  \n smartphones and tablets.  David is a VP of Technology leading a R&D team  \n focusing on processor research investigating novel circuit and architecture  \n techniques to optimize power and performance in mobile devices.  On the  \n side\, he likes to dabble in developing Android apps.  David has a MS  \n in Electrical Engineering from UCSD in California.  He has 11 patents plus  \n several pending.     
END:VEVENT
BEGIN:VEVENT
UID:calendar.10513.field_date.0.184
SUMMARY:John Duchi: SILO: The Asymptotics of Ranking Algorithms
DTSTAMP:20130618T164412Z
DTSTART:20121010T173000Z
DTEND:20121010T183000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/silo-asymptotics-ranking-algorithms
LOCATION:WID 3280B - Computer...
DESCRIPTION:  Abstract: We consider the predictive problem of supervised ranking\, where  \n the task is to rank sets of candidate items returned in response to queries.  \n Although there exist statistical procedures that come with guarantees of  \n consistency in this setting\, these procedures require that individuals  \n provide a complete ranking of all items\, which is rarely feasible in  \n practice. Instead\, individuals routinely provide partial preference  \n information\, such as pairwise comparisons of items\, and more practical  \n approaches to ranking have aimed at modeling this partial preference data  \n directly. As we show\, however\, such an approach has serious theoretical  \n shortcomings. Indeed\, we demonstrate that many commonly used  \n surrogate losses for pairwise comparison data do not yield consistency\;  \n surprisingly\, we show inconsistency even in low-noise settings. With these  \n negative results as motivation\, we present a new approach to supervised  \n ranking based on aggregation of partial preferences and develop  \n U-statistic-based empirical risk minimization procedures. We present an  \n asymptotic analysis of these new procedures\, showing that they yield  \n consistency results that parallel those available for classification. We  \n complement our theoretical results with an experiment studying the new  \n procedures in a large-scale web-ranking task. Joint work with Lester Mackey  \n and Michael Jordan.  Bio: John Duchi is a fifth-year PhD student in  \n computer science at UC Berkeley\, jointly supervised by Michael Jordan and  \n Martin Wainwright. John has also worked for several years on the research  \n team at Google Research and received the Bachelor's and Master's degrees from  \n Stanford University. He is interested in large scale statistical modeling and  \n optimization\, and has done work in distributed and stochastic optimization\,  \n ranking algorithms\, and graphical models.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10394.field_date.0.185
SUMMARY:Brian Stankiewicz and Richard Moore: 3M Visual Attention Service
DTSTAMP:20130618T164412Z
DTSTART:20121010T210000Z
DTEND:20121010T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/3m-visual-attention-service
LOCATION:1240CS
DESCRIPTION:The human visual system is a complex system that accomplishes remarkable  \n tasks in real time. The human visual system has adopted many computational  \n "tricks" to help it process in real time the constant bombardment of visual  \n information that is being presented to it. One way that the human visual  \n system deals with this massive amount of information is to select only a  \n sub-set of information available and "attend" (i.e.\, deeply process) to that  \n information.  3M research has conducted its own basic research and leveraged  \n basic research in academia to build a computer vision model that predicts  \n what people will notice in the first 3-5 seconds of viewing a scene or image.  \n  This research has been translated into a commercialized product called  \n Visual Attention Service (vas.3m.com) and deployed to Microsoft's Azure cloud  \n computing system.  We will discuss the research conducted to develop the  \n model along with the process of translating this basic research to a  \n commercialized product.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10395.field_date.0.186
SUMMARY:Brian Stankiewicz and Richard Moore: 3M Company Info Session (and pizza)
DTSTAMP:20130618T164412Z
DTSTART:20121010T220000Z
DTEND:20121010T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/3m-company-info-session-and-pizza
LOCATION:1240CS
DESCRIPTION:tba
END:VEVENT
BEGIN:VEVENT
UID:calendar.10505.field_date.0.187
SUMMARY:Sayandeep Sen: Value Aware Wireless Protocol Design
DTSTAMP:20130618T164412Z
DTSTART:20121011T143000Z
DTEND:20121011T163000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/value-aware-wireless-protocol-design
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: Suman Banerjee (advisor)\;\, Aditya Akella\, Paul Barford\, Parmesh  \n Ramanathan\, Jin Li (Microsoft Research)
END:VEVENT
BEGIN:VEVENT
UID:calendar.10499.field_date.0.188
SUMMARY:Dr. Jin Li (MSR): The Quest Towards Glitch Free Real-Time Communication
DTSTAMP:20130618T164412Z
DTSTART:20121011T210000Z
DTEND:20121011T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/quest-towards-glitch-free-real-time-communication
LOCATION:CS 1221
DESCRIPTION:Abstract: Real-Time Communication (RTC) applications are on the rise. Skype\,  \n a popular application for VoIP\, video conferencing\, file sharing and instant  \n messaging\, is growing at a rapid pace. Skype users have logged 115 billion  \n minutes in the quarter of July 2012\, which is up 50 percent compared to the  \n prior quarter and represents about one fifth of the world long distance  \n calls. 43% of Skype-to-Skype calls involve video. RemoteFX\, a key feature in  \n Windows 8\, is revolutionize desktop virtualization\, and enables unprecedented  \n Cloud Computing experience. Quality of Service has proven to be critical for  \n the commercial success and the customer satisfaction of the RTC applications.  \n In this talk\, we will discuss a number of client and platform technologies  \n that improve the Quality of Service of RTC applications. This will include:  \n 1) the use of forward error correction (FEC) in RTC applications\, 2)  \n bandwidth management\, and 3) quality of service monitoring platforms. Bio:  \n Dr. Jin Li is a Research Manager and Principal Researcher at Microsoft  \n Research (Redmond\, WA). He manages the Compression\, Communication and Storage  \n group. Blending theory and system\, Dr. Li excels at interdisciplinary  \n research\, and is dedicated to advance communication and information theory  \n and apply it to practical system building. He received his Ph.D. (with honor)  \n from Tsinghua University in 1994. After brief stints at USC and Sharp Labs\,  \n he joined Microsoft Research in 1999\, first as one of the founding members of  \n Microsoft Research Asia\, and then moved to Microsoft Research (Redmond\, WA)  \n in 2001. From 2000\, Dr. Li has also served as an Affiliated Professor in  \n Tsinghua University. Dr. Li's invention has been integrated into many  \n Microsoft products. Recently\, he and his group members have made key  \n contributions in Microsoft product line (e.g.\, RemoteFX for WAN in Windows 8\,  \n Primary Data Deduplication in Windows Server 2012\, and Local Reconstruction  \n Coding in Windows Azure Storage)\, that leads to commercial impact in the  \n order of hundreds of millions of dollars. He was awarded the prestigious  \n Microsoft Gold Star Service Award 4 times\, in 1999\, 2001\, 2006 and 2010. Dr.  \n Li was the recipient of Young Investigator Award from Visual Communication  \n and Image Processing’98 (VCIP) in 1998\, ICME 2009 Best Paper Award and  \n USENIX ATC 2012 Best Paper Award. He is/was the Associate Editor/Guest Editor  \n of IEEE Trans. On Multimedia\, Journal of Selected Area of Communication\,  \n Journal of Visual Communication and Image Representation\, P2P networking and  \n applications\, Journal of Communications. He was the General Chair of PV2009\,  \n the lead Program Chair of ICME 2011\, the Technical Program Chair of CCNC  \n 2013. He is an IEEE Fellow.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10501.field_date.0.189
SUMMARY:Laura LeGault and Nathanael Fillmore: AISEM 2012 - Meet the Dewey Lab
DTSTAMP:20130618T164412Z
DTSTART:20121011T210000Z
DTEND:20121011T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/aisem-2012-meet-dewey-lab
LOCATION:CS 4310
DESCRIPTION: AISEM series provide a unique opportunity for both new and returning  \n students to interact with groups of the UW Artificial Intelligence community  \n (including machine learning\, bioinformatics\, computer vision etc.) and hear  \n about their research. Each seminar will feature a unique group. The seminar  \n includes several short talks given by students in that group and a social  \n time which the PI will be present. Refreshments are provided.  In this first  \n seminar\, we will introduce Prof. Colin Dewey’s lab. Dewey lab focuses on  \n developing algorithms and statistical models for genomics\, especially for  \n analyzing RNA sequencing (RNA-Seq) data. Two members of the Dewey lab\, Laura  \n LeGault and Nathanael Fillmore\, will present their work.   Title: Inference  \n on Probabilistic Splice Graphs Speaker: Laura LeGault Abstract: Between the  \n increase in high-throughput RNA sequencing and the increase in de novo  \n transcriptome assembly\, methods accounting for alternative mRNA splicing have  \n become even more important. We discuss probabilistic models and associated  \n inference algorithms for estimating relative frequencies of alternative  \n transcription events and examine the practical applications of these models  \n to axolotl research.   Title: RSEM-EVAL: A Probabilistic Transcriptome  \n Assembly Evaluator Speaker: Nathanael Fillmore Abstract: In order to  \n understand a wide variety of cellular processes\, biologists find it useful to  \n study a cell's transcriptome\, i.e.\, the collection of RNA transcripts present  \n in the cell. High-throughput RNA sequencing (RNA-Seq) is a (fairly)  \n recently-developed\, powerful tool toward this end. However\, the physical  \n RNA-Seq protocol only produces short reads from a full RNA transcript\, so an  \n important early step in any RNA-Seq analysis is to assemble a collection of  \n these reads into longer sequences\, ideally full-length transcripts. If the  \n organism's genome is known\, this is relatively straightforward - but most  \n organisms' genomes are not known\, and in this case\, the assembly problem is  \n rather difficult. Several computer programs exist to produce an assembly when  \n the genome is not known\, but the assemblies produced by a pair of these  \n programs\, or even by a single program with different parameter settings\, can  \n differ substantially\, and it is often not clear how to choose between the  \n assemblies.   We have developed a transcriptome assembly scoring function  \n which can be used to choose the best assembly from a collection of candidate  \n assemblies. Our score does not require any knowledge of the organism's genome  \n or of the true set of RNA transcripts. The score is based on a probability  \n model of the process of RNA-Seq read generation and the process of ideal  \n transcriptome assembly. In the talk\, I will describe (i) our model\, (ii)  \n computational challenges associated with evaluation of the score\, (iii)  \n validation of the model and meta-evaluation of the scoring function\, and (iv)  \n preliminary experimental results on real and simulated data. I will begin  \n with background information so that the talk is accessible to a broad  \n audience of computer scientists.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10511.field_date.0.190
SUMMARY:Simon M. Lin: Found in Translation: Pharmacogenomics
DTSTAMP:20130618T164412Z
DTSTART:20121012T144500Z
DTEND:20121012T154500Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/found-translation-pharmacogenomics
LOCATION:Room 313\, Pyle Center\,...
DESCRIPTION:CIBM Retreat Keynote Presentation: Dr. Lin is the Director of the Biomedical  \n Informatics Research Center\, and the inaugural holder of the Dr. John Melski  \n Distinguished Physician/Scientist Endowment in Biomedical Informatics  \n Research at the Marshfield Clinic Research Foundation (MCRF). Dr. Lin  \n recently joined MCRF from Northwestern University Fienberg School of Medicine  \n and the Northwestern University Clinical and Translational Sciences Institute  \n where he was a Research Associate Professor of Biomedical Informatics. Prior  \n to that\, Dr. Lin was a Research Assistant Professor of Biostatistics and  \n Bioinformatics and Technical Manager of Duke Bioinformatics Shared Resources\,  \n Duke University Medical Center. Dr. Lin is known nationally for developing  \n quantitative models and innovative software to manage complex biomedical  \n datasets. “I am confident in his abilities to direct BIRC as he has proven  \n senior leadership on the strategic expansion of bioinformatics\, genetics\, and  \n health IT at the Northwestern University Robert H. Lurie Comprehensive Cancer  \n Center where he served as Associate Director of Bioinformatics and the Center  \n of Cancer Nanotechnology where he served as Director of Bioinformatics\,”  \n says Humberto Vidaillet\, MD\, Director of MCRF. Dr. Lin has recently been  \n named as US representative to the International Health Terminology Standards  \n Development Organization (IHTSDO). He is serving on the Implementation and  \n Innovation Committee for a two-year term of 2012-2013.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10510.field_date.0.191
SUMMARY:Nick Street: Data Mining for Personalized Health Care Decision Making
DTSTAMP:20130618T164412Z
DTSTART:20121012T181500Z
DTEND:20121012T191500Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/data-mining-personalized-health-care-decision-making
LOCATION:Room 313\, Pyle Center\,...
DESCRIPTION:Health care choices are made based on results of large-scale studies. The  \n care you receive is therefore based on what works for most people\, not  \n necessarily what will work for you. However\, massive datasets of patient  \n records are increasingly being mined to allow personalized health care  \n decision making. This talk will highlight work by our group that addresses  \n the following questions: How can a person choose the best lifestyle choices  \n to minimize the risk of heart disease? How can a nurse minimize a child's  \n distress during a painful medical procedure? And\, how can a physician choose  \n the best diagnostic test to reach a fast\, cheap\, and correct diagnosis?
END:VEVENT
BEGIN:VEVENT
UID:calendar.10512.field_date.0.192
SUMMARY:Patricia Flatley Brennan: Seeing Health in Every-Day Lives: Designing  \n Informatics Solutions for Healthful Patient Centered Living
DTSTAMP:20130618T164412Z
DTSTART:20121012T203000Z
DTEND:20121012T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/seeing-health-every-day-lives-designing-informatics-solutions-healthful-patient-centered-livin
DESCRIPTION:CIBM Retreat Keynote Presentation: As health care migrates from the clinic  \n and hospital to the home and community there is increasing demands for really  \n useful health information technologies (HIT) that support patient  \n self-management and engagement in clinical care. Yet most of the information  \n technologies available to patients are replicates of or extensions of tools  \n originally built for professionals in the work place. Project HealthDesign\, a  \n national program of the Robert Wood Johnson Foundation\, supported 14 teams  \n from around the country to develop innovative consumer-facing technology  \n solutions ("apps"\, devices\, data integration strategies\, etc). What emerged  \n was the unique ways patients name and label health challenges and health  \n behaviors and the importance of incorporating into design of consumer-facing  \n HIT solutions understanding WHERE health activities occur. We developed the  \n Living Environments Laboratory at the Wisconsin Institutes for Discovery\, and  \n created a virtual reality CAVE that allows us to re-create every home  \n environment on earth. The CAVE\, originally built to accelerate the design of  \n home care technologies\, now serves as a site for a wide range of scholarly  \n activities\, from art installations to visualization of molecules to  \n exploration of vision and perceptions.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10498.field_date.0.193
SUMMARY:Gilles Muller: Diagnosys: Automatic Generation of a Debugging Interface to  \n the Linux Kernel
DTSTAMP:20130618T164412Z
DTSTART:20121012T204500Z
DTEND:20121012T214500Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/diagnosys-automatic-generation-debugging-interface-linux-kernel
LOCATION:1221
DESCRIPTION:Abstract:   The Linux kernel does not export a stable\, well-defined kernel  \n interface\, complicating the development of kernel-level services\, such as  \n device drivers and file systems. While there does exist a set of functions  \n that are exported to external modules\, this set of functions frequently  \n changes\, and the functions have implicit\, ill-documented preconditions. No  \n specific debugging support is provided.   We present Diagnosys\, an approach  \n to automatically constructing a debugging interface for the Linux kernel.  \n First\, a designated kernel maintainer uses Diagnosys to identify constraints  \n on the use of the exported functions. Based on this information\, developers  \n of kernel services can then use Diagnosys to generate a debugging interface  \n specialized to their code. When a service including this interface is tested\,  \n it records information about potential problems. This information is  \n preserved following a kernel crash or hang. Our experiments show that the  \n generated debugging interface provides useful log information and incurs a  \n low performance penalty.   Speaker Bio:   Gilles Muller received the Ph.D.  \n degree in 1988 from the University of Rennes I\, and the Habilitation a  \n Diriger des Recherches degree in 1997 from the University of Rennes I.  \n  After having been a researcher at INRIA and a Professor at the Ecole des  \n Mines de Nantes\, he is currently a senior research scientist at INRIA  \n Paris-Rocquencourt. His research interests include the development of new  \n methodologies based on domain-specific languages for the structuring of OSes  \n and the development of multicore OSes.   Gilles Muller was the PC Chair of  \n EuroSys 2010 and PLOS 2010. He was involved in more than 50 program  \n committees of international workshops and conferences such as EuroSys\,  \n ASPLOS\, DSN\, SRDS\, PLOS and the EuroSys prize for the best PhD thesis.  \n  Gilles Muller has been a member of the IEEE since 1995 and was the vice  \n chair of the ACM/SIGOPS from July 2003 to July 2007.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10529.field_date.0.194
SUMMARY:Kevin Ponto: Rethinking 3D
DTSTAMP:20130618T164412Z
DTSTART:20121016T210000Z
DTEND:20121016T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/rethinking-3d
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) Seminar: Film and  \n media outlets have recently found new profitability in the use of 3D media.  \n Unfortunately\, the current means of content creation and presentation create  \n an experience that does not match the viewer's perceptual expectations. While  \n Virtual Reality systems have the potential to mitigate these shortcomings\,  \n their capabilities are often underutilized. This talk will discuss the means  \n in which humans determine depth and the methods in which stereo imagery is  \n often presented. The talk will focus on recently completed research that  \n enables the calibration of a virtual environment based on an individual's  \n perception. Empirical evaluation demonstrates that these newly developed  \n methods improve the viewer's experience by reducing many of the perceptual  \n artifacts found in traditional 3D environments.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10514.field_date.0.195
SUMMARY:Emily Provost: SILO: Emotions in Engineering: Methods for the Interpretation  \n of Ambiguous Emotional Content
DTSTAMP:20130618T164412Z
DTSTART:20121017T173000Z
DTEND:20121017T183000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/silo-emotions-engineering-methods-interpretation-ambiguous-emotional-content
LOCATION:WID 3280B - Computer...
DESCRIPTION:  Abstract Emotion has intrigued researchers for generations. This  \n fascination has permeated the engineering community\, motivating the  \n development of affective computational models for classification. However\,  \n human emotion remains notoriously difficult to interpret both because of the  \n mismatch between the emotional cue generation (the speaker) and cue  \n perception (the observer) processes and because of the presence of complex  \n emotions\, emotions that contain shades of multiple affective classes. Proper  \n representations of emotion would ameliorate this problem by introducing  \n multidimensional characterizations of the data that permit the quantification  \n and description of the varied affective components of each utterance.  My  \n work seeks to provide a computational account of how humans perceive  \n emotional utterances. I leverage perception estimation studies to develop  \n systems capable of interpreting naturalistic expressions of emotion and to  \n create novel quantification measures. This area has applications in the  \n design of affective avatars\, the development of novel machine learning  \n algorithms\, and in furthering our scientific understanding of human emotion  \n expression.  In this talk I will discuss Emotion Profiles and Emotograms\,  \n quantitative measures expressing the degree of the presence or absence of a  \n set of basic emotions within an expression. They avoid the need for a  \n hard-labeled assignment by instead providing a method for describing the  \n shades of emotion present in an utterance. These profiles can be used to  \n determine a most likely assignment for an utterance\, to map out the evolution  \n of the emotional tenor of an interaction\, or to interpret utterances that  \n have multiple affective components. The Emotion-Profile technique is able to  \n accurately identify the emotion of utterances with definable ground truths  \n (emotions with an evaluator consensus) and is able to interpret the affective  \n content of emotions with ambiguous emotional content (no evaluator  \n consensus)\, emotions that are typically discarded during classification  \n tasks. I will present results detailing the construction\, application\, and  \n benefit of this representation paradigm.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10531.field_date.0.196
SUMMARY:Lakshmikant Shrinivas and Matt Fuller: Query Optimization in the Vertica  \n Analytic Database
DTSTAMP:20130618T164412Z
DTSTART:20121017T180000Z
DTEND:20121017T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/query-optimization-vertica-analytic-database
LOCATION:CS 4310
DESCRIPTION:AbstractPlease join us as we talk about query optimization in the Vertica  \n Analytic Database. We will go over some unique challenges in query  \n optimization in column-store databases that led to our writing the optimizer  \n from scratch. We will also cover some software engineering design decisions  \n that motivated Vertica’s modular\, distributed query optimizer.About the  \n SpeakersMatt Fuller has been a developer at Vertica (now an HP company) for  \n more than 4 years\, and currently leads the Optimizer team. Lakshmikant  \n (Pachu) Shrinivas has been a developer at Vertica for more than 3 years\, and  \n was deeply involved in the query optimizer. He currently leads the Analytics  \n team.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10517.field_date.0.197
SUMMARY:Maxwell Collins\, Won Hwa Kim and Deepti Pachauri : AISEM 2012 - Meet Vikas  \n Singh Lab
DTSTAMP:20130618T164412Z
DTSTART:20121018T210000Z
DTEND:20121018T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/aisem-2012-meet-vikas-singh-lab
LOCATION:CS 4310
DESCRIPTION:AISEM series provide a unique opportunity for both new and returning students  \n to interact with groups of the UW Artificial Intelligence community  \n (including machine learning\, bioinformatics\, computer vision etc.) and hear  \n about their research. Each seminar will feature a unique group. The seminar  \n includes several short talks given by students in that group and a social  \n time which the PI will be present. **Refreshments** are provided. In this  \n seminar\, we will introduce Prof. Vikas Singh’s lab. Singh lab focuses on  \n problems motivated from image data with a distinct optimization and/or  \n geometric flavor in the field of Computer Vision\, Medical Image Analysis\, and  \n Machine Learning. Three members of this lab\, Maxwell Collins\, Won Hwa Kim and  \n Deepti Pachauri will present their work. *Title* Random Walks based  \n Multi-Image Segmentation: Quasiconvexity Results and GPU-based Solutions  \n *Speaker* Maxwell Collins *Abstract* We recast the Cosegmentation problem  \n using Random Walker (RW) segmentation as the core segmentation algorithm\,  \n rather than the traditional MRF approach adopted in the literature so far.  \n Our formulation is similar to previous approaches in the sense that it also  \n permits Cosegmentation constraints (which impose consistency between the  \n extracted objects from ≥ 2 images) using a nonparametric model. However\,  \n several previous nonparametric cosegmentation methods have the serious  \n limitation that they require adding one auxiliary node (or variable) for  \n every pair of pixels that are similar (which effectively limits such methods  \n to describing only those objects that have high entropy appearance models).  \n In contrast\, our proposed model completely eliminates this restrictive  \n dependence — the resulting improvements are quite significant. Our model  \n further allows an optimization scheme exploiting quasiconvexity for  \n model-based segmentation with no dependence on the scale of the segmented  \n foreground. Finally\, we show that the optimization can be expressed in terms  \n of linear algebra operations on sparse matrices which are easily mapped to  \n GPU architecture. We provide a highly specialized CUDA library for  \n Cosegmentation exploiting this special structure\, and report experimental  \n results showing these advantages. *Title* Wavelet based multi-scale shape  \n features on arbitrary surfaces for cortical thickness discrimination  \n *Speaker* Won Hwa Kim *Abstract* Hypothesis testing on signals defined on  \n surfaces (such as the cortical surface) is a fundamental component of a  \n variety of studies in Neuroscience. The goal here is to identify regions that  \n exhibit changes as a function of the clinical condition under study. As the  \n clinical questions of interest move towards identifying very early signs of  \n diseases\, the corresponding statistical differences at the group level  \n invariably become weaker and increasingly hard to identify. Indeed\, after a  \n multiple comparisons correction is adopted (to account for correlated  \n statistical tests over all surface points)\, very few regions may survive. In  \n contrast to hypothesis tests on point-wise measurements\, in this paper\, we  \n make the case for performing statistical analysis on multi-scale shape  \n descriptors that characterize the local topological context of the signal  \n around each surface vertex. Our descriptors are based on recent results from  \n harmonic analysis\, that show how wavelet theory extends to non-Euclidean  \n settings (i.e.\, irregular weighted graphs). We provide strong evidence that  \n these descriptors successfully pick up group-wise differences\, where  \n traditional methods either fail or yield unsatisfactory results. Other than  \n this primary application\, we show how the framework allows performing  \n cortical surface smoothing in the native space without mappint to a unit  \n sphere. *Title* Incorporating Domain Knowledge in Matching Problems via  \n Harmonic Analysis *Speaker* Deepti Pachauri *Abstract* Matching one set of  \n objects to another is a ubiquitous task in machine learning and com- puter  \n vision that often reduces to some form of the quadratic assignment problem  \n (QAP). The QAP is known to be notoriously hard\, both in theory and in  \n practice. Here\, we in- vestigate if this difficulty can be mitigated when  \n some additional piece of information is available: (a) that all QAP instances  \n of interest come from the same application\, and (b) the correct solution for  \n a set of such QAP instances is given. We propose a new approach to accelerate  \n the solution of QAPs based on learning parameters for a modified objective  \n function from prior QAP instances. A key feature of our approach is that it  \n takes ad- vantage of the algebraic structure of permutations\, in conjunction  \n with special methods for optimizing functions over the symmetric group Sn in  \n Fourier space. Experiments show that in practical domains the new method can  \n outperform existing approaches.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10457.field_date.0.198
SUMMARY:Parmesh Ramanathan\, Professor: Confidentiality-preserving Optimization in the  \n Cloud
DTSTAMP:20130618T164412Z
DTSTART:20121022T210000Z
DTEND:20121022T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/confidentiality-preserving-optimization-cloud
LOCATION:Wisconsin Institute for...
DESCRIPTION:Cloud computing paradigm offers customers instant and affordable access to a  \n distributed pool of computational resources for solving large complex  \n problems.  However\, customers are still wary of sending their problems to  \n public cloud infrastructure due to security concerns. One of their major  \n worries is that key proprietary design information about their application  \n will be comprised. In this talk\, I will discuss an approach for obfuscating  \n critical design information when solving large optimization problems in the  \n cloud. <?xml:namespace prefix = o /> In our approach\, the customer first  \n models his/her application as an optimization problem. Key design information  \n in the optimization problem is then masked using certain  \n optimality-preserving transformations. The masked optimization problem is  \n sent to the cloud and its best solution is returned.  The customer  \n re-transforms the cloud's solution to obtain the desired solution to the  \n original problem.  The challenges are to ensure that: (i) the cloud cannot  \n reconstruct key design information from the masked problem and\, (ii) the  \n final answer is best solution to the original problem. In this talk\, I will  \n illustrate our approach using two optimization problems\, one from the area of  \n electronic design automation and the other from power engineering.  
END:VEVENT
BEGIN:VEVENT
UID:calendar.10500.field_date.0.199
SUMMARY:David Andersen: How Architecture Drove Me To Submit To An Algorithms  \n Workshop: Lessons from the FAWN project
DTSTAMP:20130618T164412Z
DTSTART:20121022T210000Z
DTEND:20121022T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/how-architecture-drove-me-submit-algorithms-workshop-lessons-fawn-project
LOCATION:1221
DESCRIPTION:Abstract:   In this talk\, I'll detail a set of experiences building  \n energy-efficient clusters of "wimpy" nodes.  Using large numbers of low-GHz  \n processors to accomplish big data tasks capitalizes on several underlying\,  \n fundamental realities of computer hardware: High frequency is inefficient\;  \n easily exploitable parallelism in programs is good\; designing systems with an  \n intense focus on locality is similarly good.   It is also painful.   In the  \n past four years\, we've built a series of these "fast arrays of wimpy nodes"\,  \n containing from 20 to 85 nodes\, and used them to accomplish tasks such as  \n high-performance key-value storage using flash memory\; searching text  \n corpuses for millions of search phrases at a time\; and in-memory caching a la  \n memcached.  In each of these cases\, the changes required to achieve  \n efficient\, fast performance were substantial\, ranging from simple manual  \n optimization and configuration changes\; to software design changes\; to  \n algorithmic changes and the re-engineering of algorithms to accomplish the  \n tasks.  The engineering and programming efforts for each of these have been  \n large\, though worthwhile: The systems we constructed were several times more  \n energy-efficient than their predecessors.   Speaker Biography: David  \n Andersen is an assistant professor in the Computer Science department at  \n Carnegie Mellon University. He received his Ph.D. and M.S. degrees from MIT\,  \n and received B.S. degrees in Computer Science and Biology from the University  \n of Utah. Before joining MIT\, he was a co-founder and CTO of an Internet  \n Service Provider in Salt Lake City. His research interests center on computer  \n systems in the networked environment and energy-efficient computing.The talk  \n will conclude on a mixed note of caution and optimism.  The FAWN project\,  \n with all of its attendant challenges in software construction\, is one  \n potential harbinger of the things that architecture may deliver in the  \n future: Not just a future of parallelism\, but a future of massive  \n parallelism\, mandatory locality\, and very constrained per-node memory.  \n  Fortunately\, continuing to provide better\, faster software in such a future  \n should keep the computer science community as a whole out of trouble for a  \n few decades\, requiring simultaneous\, concerted help from architecture\,  \n theory\, and programming languages.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10448.field_date.0.200
SUMMARY:Prof Mark Hill: Why On-Chip Cache Coherence is Here to Stay
DTSTAMP:20130618T164412Z
DTSTART:20121023T210000Z
DTEND:20121023T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/why-chip-cache-coherence-here-stay
LOCATION:1240 CS
DESCRIPTION:Today’s multicore chips commonly implement shared memory with  \n cachecoherence as low-level support for operating systems and  \n applicationsoftware. Technology trends continue to enable the scaling of the  \n numberof (processor) cores per chip. Because conventional wisdom says that  \n thecoherence does not scale well to many cores\, some prognosticatorspredict  \n the end of coherence.This talk refutes this conventional wisdom by showing  \n one way to scaleon-chip cache coherence with bounded costs by combining known  \n techniquessuch as: shared caches augmented to track cached copies\, explicit  \n cacheeviction notifications\, and hierarchical design. Based upon  \n ourscalability analysis of this proof-of-concept design\, we predict  \n thaton-chip coherence and the programming convenience and compatibility  \n itprovides are here to stay.See paper by Milo M. K. Martin\, Mark D. Hill\, &  \n Daniel J. Sorin\, Comm.of the ACM\, July  \n 2012\, http://dx.doi.org/10.1145/2209249.2209269
END:VEVENT
BEGIN:VEVENT
UID:calendar.10536.field_date.0.201
SUMMARY:Maureen A. Smith: Health Services Research: Integrating Clinical\,  \n Administrative\, and Survey Data for Research on Improving Health Care
DTSTAMP:20130618T164412Z
DTSTART:20121023T210000Z
DTEND:20121023T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/health-services-research-integrating-clinical-administrative-and-survey-data-research-improvin
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) Seminar: Dr.  \n Smith¹s research program examines the effectiveness of our health care  \n system for aging and chronically ill persons. She is currently examining  \n guideline adherence and short/long-term health outcomes in patients with  \n diabetes (R01 HS018368) and the effect of paying for virtual colonoscopy on  \n colorectal cancer screening (R01 CA144835). She has over 75 peer-reviewed  \n publications in journals including JAMA\, Annals of Internal Medicine\, Medical  \n Care\, Health Services Research\, and Quality and Safety in Healthcare. Her  \n leadership positions include the Director of the UW Health Innovation  \n Program\, Director of the Community Academic Partnerships core of the NIH-CTSA  \n funded Institute for Clinical and Translational Research\, and Associate  \n Director for Population Sciences at the UW Carbone Cancer Center. Her  \n educational activities include directing an AHRQ-funded training grant in  \n health services research (T32 HS000083)\, teaching two graduate courses in  \n translating research into practice and the quality of healthcare and patient  \n safety\, and mentoring and advising clinical faculty\, fellows\, residents\,  \n medical students\, and undergraduates. In 2009\, she received the AHRQ John M.  \n Eisenberg Excellence in Mentorship Award. Dr. Smith received MD and MPH  \n degrees from Yale University School of Medicine\, and a PhD from the  \n University of Minnesota in Health Services Research\, Policy and  \n Administration in 1999.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10537.field_date.0.202
SUMMARY:Kwang-Sung Jun\, Bryan R. Gibson: AISEM: Meet the Zhu Lab
DTSTAMP:20130618T164412Z
DTSTART:20121025T210000Z
DTEND:20121025T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/aisem-meet-zhu-lab
LOCATION:4310cs
DESCRIPTION:This AISEM series provides a unique opportunity for both new and returning  \n students to interact with groups of the UW Artificial Intelligence community  \n (including machine learning\, bioinformatics\, computer vision etc.) and hear  \n about their research. Each seminar will feature a unique group. The seminar  \n includes several short talks given by students in that group and a social  \n time which the PI will be present. **Refreshments** are provided. In this  \n seminar\, we will introduce Prof. Jerry Zhu’s lab. Specific topics studied  \n in this lab include semi-supervised learning\, Bayesian nonparametrics\,  \n machine learning against bullying\, and computational models for cognitive  \n science. Two members of this lab\, Kwang-Sung Jun and Bryan R. Gibson will  \n present their work. Title: Learning from Human List Production Speaker:  \n Kwang-Sung Jun Human list production is the process of spontaneously  \n generating an ordered list of items in response to some input. It generates a  \n form of non-iid data with important applications in cognitive psychology and  \n machine learning. We propose asampling-with-discounted-replacement (SWDR)  \n model for human list production. We discuss its relation to other sampling  \n models\, and provide a parameter estimation procedure for learning. Two  \n real-world applications demonstrate the value of our model: (i) in verbal  \n fluency\, our estimated parameters align well with psychological factors  \n thought to influence behaviors in healthy and brain-damaged humans\; (ii) in  \n feature volunteering\, our model improves the accuracy of text classifiers  \n trained by Generalized Expectation criteria by learning from feature-label  \n pairs spontaneously produced by human teachers. Title: Using Machine Learning  \n to Understand and Influence Human Categorization Behavior Speaker: Bryan R.  \n Gibson In both machine learning (ML) and cognitive psychology\, categorization  \n is considered a basic task commonly encountered by learning agents as studied  \n in both fields. While a great deal of work in cognitive psychology has been  \n applied to understanding human learning in supervised categorization\, almost  \n no work has been done previously to investigate the effects of both labeled  \n and unlabeled data as in the semi-supervised paradigm. I will describe  \n several projects which use ML to both (1) better understand how labeled and  \n unlabeled data affect human learners in categorization tasks as well as (2)  \n attempt to influence the resulting behavior using ideas and techniques  \n derived from ML.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10538.field_date.0.203
SUMMARY:Aaron Gember: Toward Software-Defined Middlebox Networking
DTSTAMP:20130618T164412Z
DTSTART:20121026T193000Z
DTEND:20121026T200000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/toward-software-defined-middlebox-networking
LOCATION:CS 4310
DESCRIPTION:Current middlebox management mechanisms are clumsy and unsuitable for taking  \n full advantage of new MB deployment models and diverse middlebox  \n functionality. Instead\, we advocate for mechanisms that help exercise unified  \n control over the key factors influencing middlebox operations. Our goal is to  \n realize a software-defined middlebox networking framework to simplify  \n management of complex\, diverse functionalities and engender rich deployments.  \n In this talk\, I will discuss the major challenges that arise---representing\,  \n manipulating\, and knowledgeably controlling MB state---and the potential  \n interfaces and abstractions for addressing them.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10509.field_date.0.204
SUMMARY:Christian Kirches: Fast numerical methods for mixed-integer nonlinear model  \n predictive control
DTSTAMP:20130618T164412Z
DTSTART:20121029T210000Z
DTEND:20121029T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/fast-numerical-methods-mixed-integer-nonlinear-model-predictive-control
LOCATION:Wisconsin Institute for...
DESCRIPTION:We are interested in the fast solution of nonlinear ODE/DAE-constrained  \n mixed-integer optimal control and model predictive control problems. Such  \n problems frequently arise in industrial process control\, and typically show  \n significant potential for optimization. The hybrid and nonlinear nature of  \n these problems however is challenging to deal with.   We present a  \n computational framework based on a direct and simultaneous method for optimal  \n control and on a partial outer convexification reformulation of the problem.  \n We show how to efficiently compute approximate solutions with feasibility and  \n optimality certificates\, and can typically do so without experiencing  \n exponential runtime. The concept of real-time iterations also allows for a  \n transfer of our framework to closed-loop control. Here\, the computational  \n performance is determined by the effort required to solve one nonconvex  \n feedback QP in each real-time iteration. Block structures are exploited to  \n significantly reduce this effort. We conclude with an outlook on current  \n algorithmic developments in mixed-integer nonlinear model-predictive control.  \n                
END:VEVENT
BEGIN:VEVENT
UID:calendar.10552.field_date.0.205
SUMMARY:Amit Acharya: Computation and Informatics in Dentistry: A Specialized Area of  \n Research Focus
DTSTAMP:20130618T164412Z
DTSTART:20121030T210000Z
DTEND:20121030T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/computation-and-informatics-dentistry-specialized-area-research-focus
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) Seminar: As the  \n healthcare in United States enter the era of accountability and move towards  \n achieving the “Triple Aim”\, dental and craniofacial care should not be  \n dismissed. The cost of treating dental care is a growing concern. Based on a  \n projection from the Centers for Medicare and Medicaid Services the total  \n national expenditures for dental care will almost triple between 2000 and  \n 2020 (from $62.0 billion in 2000 to $167.9 billion in 2020\, a 271% increase.  \n As electronic health records become more widespread as a result of federal  \n mandate and along with the many recent oral and systemic health research\, the  \n feasibility of integrating medical and dental care\, research and education  \n are being explored by many healthcare organizations. The field of dental  \n informatics can play a major role in facilitating the transformation of  \n dentistry into the new era. In this talk Dr. Acharya will provide a brief  \n introduction to the field of dental informatics and where it fits within the  \n larger realm of the informatics world. He will also go over the highlights of  \n the ‘Dental Informatics Research and Training Program’ established at the  \n Marshfield Clinic Research Foundation. In closing\, Dr. Acharya will provide  \n an overview of two research studies that was conducted at his lab to gain an  \n understanding on the different flavors of research conducted in this emerging  \n field. Recently\, as part of the CIBM program\, the University of  \n Wisconsin-Madison\, partnering with Marshfield Clinic\, received funding from  \n the National Institute for Dental and Craniofacial Research (NIDCR)  \n (2T15LM007359-11) to provide three slots for dental informatics postdoctoral  \n training as part of its biomedical informatics training program. The CIBM  \n program at UW-Madison is one of the three sites in the nation to receive this  \n funding to provide dental informatics training. Oregon Health and Sciences  \n University and University of Pittsburgh were the other two sites which  \n received funding.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10502.field_date.0.206
SUMMARY:Jin-Yi Cai: Post's Problem Revisited---A Complexity Dichotomy Perspective  \n ********THIS TALK IS POSTPONED TO NEXT WEEK**********
DTSTAMP:20130618T164412Z
DTSTART:20121030T211500Z
DTEND:20121030T221500Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/posts-problem-revisited-complexity-dichotomy-perspective-talk-postponed-next-week
LOCATION:901 Van Vleck Hall
DESCRIPTION:*************This talk is postponed to next week Nov 6*********************  \n Post's Problem Revisited---A Complexity Dichotomy Perspective Jin-Yi Cai
END:VEVENT
BEGIN:VEVENT
UID:calendar.10553.field_date.0.207
SUMMARY:Radu Sion: Ancestral Myths on Clouds and Other Weather Phenomena. (or on how  \n to raise rabbits on a space ship and become rich doing it)
DTSTAMP:20130618T164412Z
DTSTART:20121102T160000Z
DTEND:20121102T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/ancestral-myths-clouds-and-other-weather-phenomena-or-how-raise-rabbits-space-ship-and-become-
LOCATION:2310 CS
DESCRIPTION:ABSTRACT: In this talk we explore the economics of cloud computing in general  \n and outsourcing your virtual machines in particular. We identify cost  \n trade-offs and postulate the key principles of outsourcing that define when  \n cloud deployment is appropriate and why. We also briefly touch on several  \n main cyber-security aspects that impact the appeal of clouds. We outline and  \n investigate some of the main research challenges on optimizing for these  \n trade-offs. If you come to this talk you are also very likely to find out  \n exactly how many US dollars you need to spend to break your favorite cipher\,  \n or send one of your bits over the network. And of course about the rabbits.  \n BIO: Radu is an Associate Professor of Computer Science at Stony Brook  \n University. He remembers when gophers were digging through the Internets and  \n bits were running at slower paces of 512 per second. He is also interested in  \n efficient computing with a touch of cyber-security paranoia\, raising rabbits  \n on space ships and sailing catamarans of the Hobie variety.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10555.field_date.0.208
SUMMARY:Aaron Strong\, Assistant Professor: General Equilibrium Ecosystem Modeling  \n with Alternative Preference Specifications
DTSTAMP:20130618T164412Z
DTSTART:20121105T220000Z
DTEND:20121105T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/general-equilibrium-ecosystem-modeling-alternative-preference-specifications
LOCATION:Wisconsin Institute for...
DESCRIPTION:Recent work by Finnoff\, Strong and Tschirhart (2008) tackles the problem of  \n how to incorporate a dynamic optimizer into the General Equilibrium Ecosystem  \n Model (GEEM). In all of the previous work that uses GEEM\, all individuals  \n within the ecosystem are net fitness maximizers. Much of the other literature  \n that models rangeland\, in particular Boyd (1991) and Huffaker and Cooper  \n (1995)\, assumes that cattle follow the Michaelis-Menten satiation grazing  \n equations. Unlike GEEM\, the Michaelis-Menten satiation equations are akin to  \n cattle having lexicographic preferences. This simple change in the  \n assumptions of cattle behavior do not allow us to directly compare the work  \n of Finnoff\, Strong and Tschirhart with that of papers such as Boyd (1991) and  \n more importantly Huffaker and Cooper (1995). This paper closes that gap  \n between these two strains of a similar literature as well as considers an  \n alternate grazing decision based on proportional grazing. Since the species  \n composition will vary depending on fitness net energy maximization and  \n lexicographic preference maximization\, we may obtain different steady states  \n or transition paths. This work identifies the key potential drivers that have  \n been consider for invasion but also identifies an alternative path way that  \n previous literature has failed to recognize. Finally\, we consider a simple  \n experiment using remote sensing data from Boulder County\, Colorado to  \n potentially eliminate some of the reasons for invasion.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10559.field_date.0.209
SUMMARY:Chen-Han Ho: Dynamic Hardware Specialization: Microarchitecture\,  \n Implementation and Evaluation
DTSTAMP:20130618T164412Z
DTSTART:20121106T160000Z
DTEND:20121106T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/dynamic-hardware-specialization-microarchitecture-implementation-and-evaluation
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: Karu Sankaralingam (advisor)\, Guri Sohi\, Mark HIll\, Nam Sung Kim
END:VEVENT
BEGIN:VEVENT
UID:calendar.10561.field_date.0.210
SUMMARY:Takeo Kanade: First-Person Vision
DTSTAMP:20130618T164412Z
DTSTART:20121106T213000Z
DTEND:20121106T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/first-person-vision
LOCATION:1240
DESCRIPTION:For understanding the behavior\, intent\, and environment of a person\, the  \n surveillance metaphor is traditional\; that is\, install cameras in the  \n environment and observe her and her interaction with other people and  \n environment from them. Instead\, we argue that the First-Person Vision that  \n senses the environment and her activities from her point of view is more  \n advantageous with available images about her environment from her own view  \n points and with readily available information about her head motion and gaze.  \n We have been working in this paradigm for a while\, and this talk will present  \n the current progresses in the First Person Vision – the ideas\, devices\,  \n algorithms\, and example applications.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10562.field_date.0.211
SUMMARY:Jin-Yi Cai: Post's Problem Revisited---A Complexity Dichotomy Perspective
DTSTAMP:20130618T164412Z
DTSTART:20121106T221500Z
DTEND:20121106T231500Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/posts-problem-revisited-complexity-dichotomy-perspective
LOCATION:901 Van Vleck Hall
DESCRIPTION:Post's Problem Revisited---A Complexity Dichotomy Perspective Jin-Yi Cai Emil  \n Post initiated the investigation regarding the existence of r.e. Turing  \n degrees between 0 and 0'. The famous solution came independently from Richard  \n Friedberg and Albert Muchnik\, who invented the finite injury priority  \n argument. The rest\, as they say\, is history---Recursion Theory has blossomed  \n into a formidable mathematical discipline. Complexity Theory owes greatly to  \n Recursion Theory in its early days. Many basic concepts and proof techniques  \n were borrowed directly from it. However\, recently the two subjects have  \n diverged. I would like to revisit Post's Problem. I will report on a  \n substantial new theory which has the philosophical underpinning that\, at  \n least in Complexity Theory\, maybe an anti-Friedberg-Muchnik theory is the  \n right answer. We will present theorems which have the general flavor that\, in  \n a broad class of relevant computational problems\, there are really only two  \n levels of problem complexity: one is tractable\, the other is complete. I will  \n survey a number of recent dichotomy theorems that achieve this complete  \n classification\, in the setting of counting problems. Specifically I will  \n discuss (1) Graph Homomorphisms\, (2) Counting Constraint Satisfaction  \n Problems\, and (3) Holant Problems. http://arxiv.org/abs/0903.4728  \n http://arxiv.org/abs/1008.0683 http://arxiv.org/abs/1204.6445 One speculation  \n is that there might be a parallel theory one can develop within Recursion  \n Theory. Is it time that Complexity Theory starts to pay back some debt it  \n owes to Recursion Theory? http://www.math.wisc.edu/~lempp/conf/swlc.html#cai
END:VEVENT
BEGIN:VEVENT
UID:calendar.10571.field_date.0.212
SUMMARY:Marco Molinaro: The Geometry of Online Packing Linear Programs
DTSTAMP:20130618T164412Z
DTSTART:20121108T220000Z
DTEND:20121108T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/geometry-online-packing-linear-programs
LOCATION:CS 4310
DESCRIPTION:We consider packing linear programs with m rows where all constraint  \n coefficients are normalized to be in the unit interval. The n columns arrive  \n in random order and the goal is to set the corresponding decision variables  \n irrevocably when they arrive to obtain a feasible solution maximizing the  \n expected reward. Previous (1 - epsilon)-competitive algorithms require the  \n right-hand side of the LP to be Omega((m/epsilon^2) log (n/epsilon)\, a bound  \n that worsens with the number of columns and rows. However\, the dependence on  \n the number of columns is not required in the single-row case and known lower  \n bounds for the general case are also independent of n. Our goal is to  \n understand whether the dependence on n is required in the multi-row case\,  \n making it fundamentally harder than the single-row version. We refute this by  \n exhibiting an algorithm which is (1 - \epsilon)-competitive as long as the  \n right-hand sides are Omega((m/epsilon)^2 log (m/epsilon)). Our techniques  \n refine previous PAC-learning based approaches which interpret the online  \n decisions as linear classifications of the columns based on sampled dual  \n prices. The key ingredient of our improvement comes from a non-standard  \n covering argument together with the realization that only when the columns of  \n the LP belong to few 1-d subspaces we can obtain small such covers\; bounding  \n the size of the cover constructed also relies on the geometry of linear  \n classifiers. General packing LP's are handled by perturbing the input  \n columns\, which can be seen as making the learning problem more robust.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10574.field_date.0.213
SUMMARY:Arvind Krishnamurthy: The Power of Consensus in Building Robust Distributed  \n Systems
DTSTAMP:20130618T164412Z
DTSTART:20121108T220000Z
DTEND:20121108T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/power-consensus-building-robust-distributed-systems
LOCATION:CS 1240
DESCRIPTION:Abstract: Distributed systems have traditionally favored responsiveness over  \n consistency. For example\, routing protocols apply received updates  \n immediately even though such responsiveness comes at the cost of routing  \n loops and blackholes\, middleboxes manage and update routing state without  \n making the updates durable\, and distributed storage systems trade off  \n consistency to achieve greater availability. Our position is that consistent  \n state in a distributed system not only makes its behavior more predictable\,  \n but surprisingly also improves its availability and performance. We will  \n illustrate this using three case studies in this talk: Consensus Routing -- a  \n consistency-first approach that cleanly separates safety and liveness\, ETTM  \n -- a new scalable and fault tolerant network manager that enables  \n participating end-hosts to cooperatively manage network resources. and  \n Scatter -- a scalable and distributed key-value storage system that provides  \n linearizable consistency even under adverse circumstances.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10581.field_date.0.214
SUMMARY:Amund Kvalbein: Resilient Networks - Measuring Mobile Broadband in Norway
DTSTAMP:20130618T164412Z
DTSTART:20121112T153000Z
DTEND:20121112T164500Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/resilient-networks-measuring-mobile-broadband-norway
LOCATION:CS 1240
DESCRIPTION:(This talk will be part of the course lecture of CS/ECE 707 which will be  \n held in the CS 1240 room. Everyone is invited to attend.) This talk will give  \n an introduction to the Resilient Networks project with a focus on our efforts  \n in measuring and improving mobile broadband performance. Mobile broadband is  \n rapidly becoming a critical service that many users are relying on every day.  \n Several events in Norway and other countries over the last years have shown  \n that failures in these networks have severe consequences. These large-scale  \n failures are\, however\, not the only source of frustration for users. Our  \n measurements show that most mobile broadband connections also experience  \n frequent short-lived periods of lost connectivity. There is a need for more  \n systematic information about the reliability and stability of mobile  \n broadband services as experienced by the end user. This information is  \n critical for businesses\, organizations and emergency services who rely on  \n these networks for performing their core tasks\, and for operators\, regulators  \n and consumers who are interested in the overall performance of mobile  \n networks. In this talk\, I will describe finished and on-going efforts in the  \n Resilient Networks project to measure the robustness of mobile broadband  \n services in Norway. Our first efforts in this area focused on measuring the  \n reliability of mobile broadband connections from a set of voting locations  \n spread across Norway. I will go through some of our findings\, and comment on  \n our experiences with performing this type of measurements. I will then talk  \n about our ongoing efforts in establishing a more permanent infrastructure for  \n monitoring the quality and robustness of Norwegian mobile broadband networks.  \n Bio: Amund Kvalbein is a Senior Research Scientist at Simula Research  \n Laboratory in Oslo\, Norway. He holds a PhD degree from the University of Oslo  \n (2007). After finishing his PhD\, he spent one year as a post doc at Georgia  \n Institute of Technology\, before returning to Oslo and Simula. He is currently  \n leader of the Resilient Networks project\, focusing on methods for improving  \n the user-experienced stability and reliability of fixed and cellular  \n communication networks. His main research interest is in the robustness and  \n performance of networks and networked services\, with a particular focus on  \n recovery and scalability at the routing layer.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10587.field_date.0.215
SUMMARY:Aaron Gember: Obtaining In-Context Measurements of Cellular Network  \n Performance
DTSTAMP:20130618T164412Z
DTSTART:20121112T200000Z
DTEND:20121112T203000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/obtaining-context-measurements-cellular-network-performance
LOCATION:CS 4310
DESCRIPTION:Network service providers\, and other parties\, require an accurate  \n understanding of the performance cellular networks deliver to users. In  \n particular\, they often seek a measure of the network performance users  \n experience solely when they are interacting with their device--a measure we  \n call in-context. Acquiring such measures is challenging due to the many  \n factors\, including time and physical context\, that influence cellular network  \n performance. We make two contributions to address the issue of obtaining  \n in-context measurements of cellular network performance. First\, we conduct a  \n large scale measurement study\, based on data collected from a large cellular  \n provider and from hundreds of controlled experiments\, to shed light on the  \n issues underlying in-context measurements. Our novel observations show that  \n measurements must be conducted on devices which (i) recently used the network  \n as a result of user interaction with the device\, (ii) remain in the same  \n macro-environment (e.g.\, indoors and stationary)\, and in some cases the same  \n micro-environment (e.g.\, in the user’s hand)\, during the period between  \n normal usage and a subsequent measurement\, and (iii) are currently  \n sending/receiving little or no user-generated traffic. Second\, we design and  \n deploy a prototype active measurement service for Android phones based on  \n these key insights. Our analysis of 1650 measurements gathered from 12  \n volunteer devices shows that the system is able to obtain average throughput  \n measurements that accurately quantify the performance experienced during  \n times of active device and network usage.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10554.field_date.0.216
SUMMARY:John Birge\, Jerry W. and Carol Lee Levin Professor of Operations Management:  \n Inferring Structure from Optimization Model Solutions
DTSTAMP:20130618T164412Z
DTSTART:20121112T220000Z
DTEND:20121112T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/inferring-structure-optimization-model-solutions
LOCATION:Wisconsin Institute for...
DESCRIPTION:Understanding the behavior of participants in markets\, such as those for  \n electricity\, requires knowledge of their views on the impact of their supply  \n (or demand) quantities. This view is in turn informed by observable prices  \n and quantities in equilibrium. For electricity markets\, system operators  \n collect and make available these data which are derived from the solution of  \n an optimization model. A key element\, the transmission network\, is\, however\,  \n not revealed. This talk will discuss the inverse optimization problem to find  \n the hidden network structure. Some results for the Midwest Independent  \n Transmission System Operator (MISO) will be presented.  \n http://www.chicagobooth.edu/faculty/bio.aspx?person_id=12824569856
END:VEVENT
BEGIN:VEVENT
UID:calendar.10582.field_date.0.217
SUMMARY:WACM Explains...What comes next? Grad School or Industry?
DTSTAMP:20130618T164412Z
DTSTART:20121112T233000Z
DTEND:20121113T003000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/wacm-explainswhat-comes-next-grad-school-or-industry
LOCATION:1221 CS
DESCRIPTION:WACM is excited to announce the first installment of WACM Explains...\, a  \n series of panel discussions on topics of interest to computer science  \n undergraduates and graduate students. Our first topic will be the differences  \n between grad school and industry\, and the advantages and disadvantages of  \n each. We will have several panelists who have experience with both grad  \n school and industry\, and we will try to provide a balanced look at the  \n choices. Not sure what to do after you finish your undergraduate degree? Want  \n to hear about some of the differences between academia and the real world?  \n Our guest panelists will be Polina Dudnik (now at Google)\, Irene Rae (current  \n PhD Student)\, and Elizabeth Soechting (now at Epic) Although this event is  \n hosted by WACM\, it is open to everyone and we encourage anyone who is  \n interested to attend. Panelist bios: Polina Dudnik: Undergraduate in Computer  \n Engineering from Binghamton University. Masters in CS from University of  \n Wisconsin working in Computer Architecture\, advised by Mark Hill. 2 years  \n working at Google\, Madison in the Platforms division. Previous internships at  \n Microsoft Research and GE Healthcare. Irene Rae: Undergraduate in Industrial  \n Design from Rochester Institute of Technology. Masters in CS from University  \n of Wisconsin working in Human-Computer Interaction\, advised by Bilge Mutlu.  \n Worked for 5 years as a freelance web developer and designer\, then another 4  \n years on a small in-house information technology team as a web administrator\,  \n developer\, and designer in an astrophysics research project. Previous  \n internships at Willow Garage\, Inc.\, a small robotics research company in  \n Silicon Valley. After 9 years working\, is now at Madison pursuing a PhD in  \n HCI. Elizabeth Soechting: Undergraduate in Computer Science from the  \n University of Virginia working in Programming Languages and Software  \n Engineering\, advised by Wes Weimer. Masters in Computer Science from the  \n University of Wisconsin. 3 years working for Epic (including during masters)  \n on electronic health records. Previous internships at Epic and Cisco.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10583.field_date.0.218
SUMMARY:Chris Hinrichs: How Machine Learning Methods Can Reshape Neuroimaging-Based  \n Clinical Trials
DTSTAMP:20130618T164412Z
DTSTART:20121113T220000Z
DTEND:20121113T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/how-machine-learning-methods-can-reshape-neuroimaging-based-clinical-trials
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) Seminar:  \n Currently\, there are no known effective treatments for Alzheimer's Disease  \n which go beyond simply delaying the onset of clinically diagnosable dementia.  \n An increasing number of trials are planned or underway\, (see  \n http://www.clinicaltrials.gov)\, but many are limited by the sensitivity and  \n power of their primary end-points\, i.e.\, statistical measures of dementia or  \n neural atrophy. Neuropsychological measures of cognitive status suffer from a  \n high degree of variability\, which can mask genuine treatment effects. This in  \n turn may require studies to recruit large patient cohorts - up to several  \n thousand - in order to have an acceptable chance of detecting differing  \n outcomes between treatment and placebo arms. The switch to neuroimaging-based  \n outcome measures has to some extent improved on this situation\, as MRI or PET  \n imaging gives a more direct means of measuring accumulating tissue damage\, or  \n hypo-activity. Yet\, this raises questions of interpretability\, and of  \n statistical efficiency: if we do detect a significant treatment-related  \n change in a patient's neuro-imaging\, does it appear to be beneficial\, and\,  \n how do we even quantify such changes? In this talk I will present a clinical  \n trial design which uses machine learning methods to determine whether or not  \n a treatment is effective\, and\, to localize these effects in order to aid  \n interpretation. Simulated clinical trials using patient MRI scans suggest  \n that this methodology can improve sensitivity by up to several orders of  \n magnitude..
END:VEVENT
BEGIN:VEVENT
UID:calendar.10575.field_date.0.219
SUMMARY:Lakshmi Bairavasundaram: NetApp Innovation: Through the Academic Lens
DTSTAMP:20130618T164412Z
DTSTART:20121114T183000Z
DTEND:20121114T193000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/netapp-innovation-through-academic-lens
LOCATION:4310 CS
DESCRIPTION:Abstract: This talk provides an introduction to NetApp and it Advanced  \n Technology Group through the lens of peer-reviewed publications. NetApp was  \n founded in 1992 by Dave Hitz\, James Lau\, and Michael Malcolm (it was called  \n Network Appliance at the time). Parts of the product were inspired by various  \n academic publications and their interplay – including NFS\, RAID\, and  \n log-structured file systems. NetApp\, therefore\, recognized the importance of  \n collaborating with academia and contributing back to academia early on. Even  \n as a startup\, the founders published a paper on WAFL\, the file system they  \n created for NetApp’s storage appliance. Over the years\, NetApp has advanced  \n the state-of-the-art in building fast\, simple\, reliable\, and efficient  \n storage systems. We’ve also published papers on all of these topics and the  \n talk will cover some of the papers “tapas”-style\, describing the gist of  \n each paper. Towards the end\, we’ll also talk about the important research  \n questions that we are working to solve\, and would love to collaborate with  \n academia on. Bio: Lakshmi N. Bairavasundaram is a member of technical staff  \n in the Advanced Technology Group at NetApp. His research interests include  \n storage systems\, file systems\, storage and data management\, and fault  \n tolerance. Lakshmi currently focuses on storage and data management  \n techniques. He joined NetApp after completing his Ph.D. in Computer Sciences  \n (2008) at the University of Wisconsin-Madison under the supervision of Prof.  \n Andrea Arpaci-Dusseau and Prof. Remzi Arpaci-Dusseau.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10585.field_date.0.220
SUMMARY:Swami Sundararaman: Snapshots in a Flash with ioSnap
DTSTAMP:20130618T164412Z
DTSTART:20121114T210000Z
DTEND:20121114T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/snapshots-flash-iosnap
LOCATION:CS 4310
DESCRIPTION:Abstract: We present ioSnap\, a system that adds snapshot capabilities to the  \n Fusion-io Virtual Sorage Layer (VSL). Through careful design and novel  \n machinery\, ioSnap delivers low-overhead snapshots with minimal disruption to  \n foreground I/O traffic. Thorough our evaluation\, we show that ioSnap incurs  \n negligible performance overhead during normal operation\, and that common-case  \n operations such as snapshot creation and deletion incur little cost. We also  \n demonstrate techniques to mitigate the performance impact on foreground I/O  \n during more intensive snapshot operations such as activation. Our prototype  \n ioSnap represents a case study of how to integrate snapshots into a modern\,  \n well-engineered flash-based storage system. Bio: Swaminathan Sundararaman is  \n Research Scientist at Fusion-io\, where he works on innovation in non-volatile  \n memory technologies and applications. Swami earned his PhD at UW Madison  \n under the guidance of Professors Andrea Arpaci-Dusseau and Remzi  \n Arpaci-Dusseau\, where he did research on building robust\, reliable\, and smart  \n systems. His work on Membrane and N-Version file systems won the best paper  \n awards at FAST '10 and USENIX ATC '09\, respectively. Swami obtained his MS  \n degree in computer science from Stony Brook University and his B.E. degree in  \n Computer Science and M.Sc. degree in Chemistry from Birla Institute of  \n Technology and Science (BITS) pilani\, India.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10586.field_date.0.221
SUMMARY:Kadir Ozdemir: Thinking Big: An EMC Tech Talk
DTSTAMP:20130618T164412Z
DTSTART:20121115T230000Z
DTEND:20121116T010000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/thinking-big-emc-tech-talk
LOCATION:CS 2310
DESCRIPTION:Data Domain (DD)\, acquired by EMC in 2009\, is one of the most innovative  \n Silicon Valley companies\, and has created the deduplication disk based backup  \n appliance market. With its inline deduplication file system\, unique data  \n invulnerability/integrity architecture\, and very efficient WAN replication  \n technologies\, DD has been the leader in the backup and recovery appliance  \n market. In this talk\, Dr. Kadir Ozdemir\, Chief Architect\, will introduce DD\,  \n the backup and recovery uses cases\, DD products and technologies\, and the  \n challenging problems keeping DD engineers busy inventing. ABOUT THE SPEAKER:  \n Kadir Ozdemir is Chief Architect for Data Domain products at EMC. Prior to  \n joining EMC\, he was Chief Architect for Veritas Storage Foundation and High  \n Availability products at Symantec. His storage industry experience covers  \n deduplication file systems\, highly available clustered file and block storage  \n systems\, and replication\, snapshot and RAID technologies. He holds a PhD  \n degree in Computer Science from University of Ottawa\, Canada.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10589.field_date.0.222
SUMMARY:Tushar Khot: Efficient Learning of Statistical Relational Models
DTSTAMP:20130618T164412Z
DTSTART:20121119T153000Z
DTEND:20121119T173000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/efficient-learning-statistical-relational-models
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: Jude Shavlik (advisor)\, Sriraam Natarajan\, David Page\, Chris Re
END:VEVENT
BEGIN:VEVENT
UID:calendar.10598.field_date.0.223
SUMMARY:Jeremy Weiss: Statistical Timeline Analysis in Relational Domains: Improving  \n Medical Prediction from Electronic Health Records
DTSTAMP:20130618T164412Z
DTSTART:20121119T210000Z
DTEND:20121119T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/statistical-timeline-analysis-relational-domains-improving-medical-prediction-electronic-healt
LOCATION:4765 Medical Sciences...
DESCRIPTION:Committee: David Page (advisor)\, Sriram Natarajan\, Mark Craven\, Deane Mosher
END:VEVENT
BEGIN:VEVENT
UID:calendar.10577.field_date.0.224
SUMMARY:Christopher DeMarco\, Professor: Optimization Problems Motivated by  \n Hamiltonian Structure in Power System Dynamics
DTSTAMP:20130618T164412Z
DTSTART:20121119T220000Z
DTEND:20121119T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/optimization-problems-motivated-hamiltonian-structure-power-system-dynamics
LOCATION:Wisconsin Institute for...
DESCRIPTION:There exists a long literature on construction of Lyapunov functions to  \n estimate basins of attraction associated with stable operating points of a  \n synchronous electric power grid. We will illustrate an interpretation of  \n these Lyapunov functions as solving an associated optimal control/variational  \n problem\, which describes the worst case\, smallest "size of disturbance" that  \n can drive the system unstable. In simplified models that possess what is  \n often termed a "nearly" Hamiltonian structure\, the associated optimal control  \n problem admits a closed form solution for the cost of control\, as a function  \n of state variables. This solution is closely related to physical stored  \n energy in the network\, and equals the traditionally derived Lyapunov  \n function. When the model admits such a closed form Lyapunov function\, one  \n useful class of optimization problem is that of computing a set of lowest  \n energy saddle exit points\, which characterize the "easiest" paths by which  \n the system may lose stability. A more challenging class of problems\, to date  \n largely unaddressed in the literature\, lies in relaxing the constraints on  \n model structure that enable the closed form solution. Here one seeks  \n computational tractable means of characterizing cost of control in the  \n optimal control problem\, for a more detailed dynamical model of the grid.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10597.field_date.0.225
SUMMARY:David M. Devilbiss: Lost Moments in Time: The Effects of Stress on PFC Neural  \n Coding
DTSTAMP:20130618T164412Z
DTSTART:20121120T220000Z
DTEND:20121120T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/lost-moments-time-effects-stress-pfc-neural-coding
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) Seminar: The  \n prefrontal cortex (PFC) plays a central role in a diverse set of cognitive  \n and behavioral processes\, including sustained attention\, working memory\, and  \n behavioral inhibition. In delayed response tasks that probe working memory  \n and other PFC functions\, fluctuations in spiking activity rates of PFC  \n neurons are posited to reflect the maintenance of attentional processes\,  \n abstract rules\, or past stimuli and events during delayed-response tasks of  \n working memory. However very little is known about coding mechanisms that  \n maintain this information for long periods of time across the delay period.  \n Additionally\, stress impairs higher cognitive processes dependent on the PFC.  \n However\, surprisingly\, to date the actions of stress on PFC neuronal  \n discharge in animals engaged in tasks of working memory also remain unknown.  \n In this talk\, I will present neurophysiologic data recorded from the PFC of  \n rats performing a delayed-response task of spatial working memory that  \n refutes the predominant theories related to stress-induced cognitive  \n impairment. In addition\, I will present a model of the conditional intensity  \n of neuronal spiking within the PFC that allows these neurons to maintain  \n information for extended periods of time\; spiking history predicted discharge  \n (SHPD). This model is framed within the context of a generalized linear model  \n to address several questions related to the actions of PFC neuron function  \n and stress. First\, to what degree does past spiking activity of PFC neurons  \n predict or modulate ongoing activity of these neurons? Second\, does stress  \n have an overall impact on the predictability of PFC neuron discharge given a  \n cell’s intrinsic spiking history? Third\, do specific task components (i.e.  \n delay-period vs. behavioral response) interact with or modulate SHPD during  \n baseline and acute stress.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10608.field_date.0.226
SUMMARY:AI Seminar: Meet the Craven Lab
DTSTAMP:20130618T164412Z
DTSTART:20121128T220000Z
DTEND:20121128T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/ai-seminar-meet-craven-lab
LOCATION:CS 3310
DESCRIPTION:In this installment of the Artificial Intelligence Seminar Series\, we will  \n hear short talks from two students from Prof. Mark Craven's lab. The group  \n develops and applies machine learning to biomedical problems\; for example\,  \n extracting structured information from scientific literature\, inferring and  \n modelling gene interaction networks\, and modelling\, classifying\, and aligning  \n temporal gene expression data. After the talks\, we'll have time for informal  \n discussion with Prof. Craven and the speakers\, Deborah Muganda-Rippchen and  \n Deborah Chasman. Refreshments are provided! Deborah Muganda-Rippchen:  \n Computing Multi-Level Clustered Alignments of Gene-Expression Time Series  \n Identifying similarities and differences in expression patterns across  \n multiple time series can provide a better understanding of the relationships  \n among various normal biological and experimentally induced conditions such as  \n chemical treatments or the effects induced by a gene knockout/suppression. We  \n consider the task of identifying sets of genes that have a high degree of  \n similarity both in their (i) expression profiles within each condition\, and  \n (ii) changes in expression responses across conditions. Previously\, we  \n developed an approach for aligning time series that computes clustered  \n alignments. In this approach\, an alignment represents the correspondences  \n between two gene expression time series. Portions of one of the time series  \n may be compressed or stretched to maximize the similarities between the two  \n series. A clustered alignment groups genes such that the genes within a  \n cluster share a common alignment\, but each cluster is aligned independently  \n of the others. Unlike standard gene-expression clustering\, which groups genes  \n according to the similarity of their expression profiles\, the  \n clustered-alignment approach clusters together genes that have similar  \n changes in expression responses across treatments. We have now extended the  \n clustered alignment approach to produce multi-level clusterings that identify  \n subsets of genes that have a high degree of similarity both in their (i)  \n expression profiles within each treatment\, and (ii) changes in expression  \n responses across treatments. We examine the validity of this multi-level  \n clustering method by performing a GO-term enrichment analysis of the  \n clusters. Additionally\, we use permutation testing to determine if our  \n clusters have alignment scores that are unlikely to occur by chance. Deborah  \n Chasman: Inferring host subnetworks involved in viral replication Viruses  \n require host machinery to complete nearly every step of their life cycle. An  \n understanding of the interactions between viruses and their hosts is needed  \n for the development of antiviral therapeutics that target specific processes.  \n High-throughput\, genome-wide screens can identify host genes involved in  \n viral replication\; however\, these screens generally produce a large amount of  \n unstructured data\, which can be difficult to interpret. We propose a method  \n to hypothesize which specific host pathways and interactions are relevant to  \n viral replication. We view the gene interaction network as a graph\, in which  \n proteins or genes are represented as nodes\, and various pairwise interactions  \n identified by experimental techniques are represented as edges. We use a  \n mixed integer linear programming approach to identify subnetworks that are  \n consistent with the biological data. The predicted subnetworks distill  \n available knowledge into a comprehensible format\, make predictions about  \n which host factors are most proximal to direct interactions with viral  \n components\, and may be used to guide further biological research.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10556.field_date.0.227
SUMMARY:Chris Harrison: HCI Seminar: Interacting with Small Devices in Big Ways
DTSTAMP:20130618T164412Z
DTSTART:20121129T220000Z
DTEND:20121129T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/hci-seminar-chris-harrison
LOCATION:1240 CS
DESCRIPTION: Interacting with Small Devices in Big Ways Despite their small size\, mobile  \n devices are able to perform tasks of creation\, information and communication  \n with unprecedented ease. However\, diminutive screens and buttons mar the  \n user experience\, and otherwise prevent us from realizing the full potential  \n of computing on the go. For example\, there is large disparity between  \n multitouch input and the capabilities of our hands and fingers. In addition  \n to translating to an X/Y position\, our fingers can vary their angle of  \n attack\, bend\, twist\, and apply different pressure and shear forces (at  \n least six additional analog dimensions). Fingers also have many “modes”  \n – they do not just poke\, as contemporary touchscreen interaction would  \n suggest\, but also scratch\, flick\, knock\, rub\, and grasp\, to name a few. I  \n will describe several technologies I have worked on that enrich and expand  \n today's interaction. I will also highlight an emergent shift in computing:  \n from mobile devices we carry to using the human body itself as an mobile  \n interactive platform\, bringing computational power ever closer to users.  \n This evolution brings significant new challenges in sensing and interaction  \n design: the human body is not only incredibly irregular and dynamic\, but also  \n comes in more than six billion different models. However\, along with these  \n challenges also come exciting new opportunities for more powerful\, intuitive  \n and intimate computing experiences. Bio: Chris Harrison is a Ph.D. candidate  \n in the Human-Computer Interaction Institute at Carnegie Mellon University. He  \n broadly investigates novel sensing technologies and interaction  \n techniques\, especially those that empower people to interact with “small  \n devices in big ways.” Harrison was recently named as one of the top 30  \n scientists under 30 by Forbes and a top 35 innovator under 35 by MIT  \n Technology Review. During his graduate studies\, Harrison has worked at  \n Microsoft Research\, IBM Research\, AT&T Labs and Disney Research on a variety  \n of topics\, from social television to on-body computing.Host: Bilge Mutlu
END:VEVENT
BEGIN:VEVENT
UID:calendar.10617.field_date.0.228
SUMMARY:Jae Young Do: The Role of Flash Memory In Database Management Systems
DTSTAMP:20130618T164412Z
DTSTART:20121130T140000Z
DTEND:20121130T160000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/role-flash-memory-database-management-systems
LOCATION:3310 CS
DESCRIPTION:Committee: Jignesh Patel (advisor)\, David DeWitt\, Jeffrey Naughton\, Michael  \n Swift\, Rafael Lazimy
END:VEVENT
BEGIN:VEVENT
UID:calendar.10557.field_date.0.229
SUMMARY:Amy Ogan: HCI Seminar: Taking Educational Technology Worldwide: Challenging  \n the Assumptions of Personalized Learning
DTSTAMP:20130618T164412Z
DTSTART:20121130T180000Z
DTEND:20121130T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/hci-seminar-amy-ogan
LOCATION:1240 CS
DESCRIPTION: Taking Educational Technology Worldwide: Challenging the Assumptions of  \n Personalized LearningThe advent of widespread access to computing offers the  \n promise to transform educational practices worldwide. At present\, most  \n massive open online courses (MOOCS) present practice and evaluation  \n opportunities through static lists of multiple-choice questions. Future  \n systems\, however\, will increasingly rely on sophisticated modeling of student  \n knowledge to provide personalized instruction. Unfortunately\, the few systems  \n that currently employ such models tend to be developed for and evaluated in  \n middle-class US schools – a very particular cultural context. This  \n assumption is a broad generalization\, as the cultural makeup of the US\, and  \n indeed the world\, is changing dramatically. A central question educational  \n technologies must confront is how these systems scale to diverse contexts  \n with differing classroom practices and values.To investigate the cultural  \n implications of educational technology use\, I studied a math tutoring system  \n that has been shown to be effective in the US. Through observation\,  \n interviews\, learning assessments\, and log data\, I explored student and  \n teacher use of the technology in school sites in Mexico\, Costa Rica\, Brazil\,  \n Portugal and Belgium with an international and local team of researchers. An  \n example finding was the much greater propensity of students in our Latin  \n American sites to collaborate closely\, engaging in interdependently-paced  \n work and very frequently conducting work away from their own computers. This  \n threatens the very core assumption of personalized learning systems - that  \n learning is individual – and rendered the sophisticated\, personalized  \n models inaccurate.The current hope is that as these systems go mainstream\,  \n the massive amounts of data collected will help researchers bootstrap and  \n improve these models. However\, the models built can only be as good as our  \n assumptions about the data. In order to achieve the full benefits of  \n personalized learning with MOOCs\, we must employ human-computer interaction  \n methodologies to help us rethink the underlying expectations for data  \n collection and system design. For instance\, our studies suggest that  \n incorporating low-cost sensors can aid systems in better determining the  \n context of use and active users. Technology designers can then implement  \n multi-user versions of their personalization algorithms (e.g.\,  \n knowledge-tracing\, help-seeking\, and adaptive scaffolding) that take  \n advantage of this contextual understanding. As technologists increasingly  \n become the gatekeepers of widespread education\, further research is warranted  \n that examines our assumptions about student identity\, classroom practices\,  \n and cultural contexts of use.Bio: Amy Ogan is currently a postdoctoral fellow  \n in Human-Computer Interaction at Carnegie Mellon University\, where she leads  \n projects investigating the development of culturally-aware educational  \n technologies that engage learners in social relationships. She completed her  \n doctorate in HCI at CMU in 2011 as the recipient of an Institute of Education  \n Sciences fellowship. Dr. Ogan has additionally been a visiting researcher at  \n USC’s Institute for Creative Technologies\, and has recently conducted field  \n research on the deployment of educational technology in seven countries. Her  \n broader research goals include the support of underprivileged students  \n through technology\, and the use of technology to introduce students to new  \n cultural perspectives. 
END:VEVENT
BEGIN:VEVENT
UID:calendar.10609.field_date.0.230
SUMMARY:Christina Leslie: Deciphering co- and post-transcriptional regulatory  \n programs in normal and cancer cells
DTSTAMP:20130618T164412Z
DTSTART:20121130T180000Z
DTEND:20121130T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/deciphering-co-and-post-transcriptional-regulatory-programs-normal-and-cancer-cells
LOCATION:140 Bardeen\, 1215 Linden...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) Seminar: Our group  \n develops computational methods to study and model transcriptional\,  \n co-transcriptional\, and post-transcriptional gene regulatory mechanisms and  \n to dissect the dysregulation of gene expression in cancer. The techniques we  \n use generally come from machine learning\, a field at the intersection of  \n computer science and statistics. In particular\, we develop supervised  \n learning methods -- algorithmic approaches for learning predictive models  \n from data -- that we train on high-throughput experimental data\, increasingly  \n from next generation sequencing. As we try to advance the state-of-the-art in  \n statistical and machine learning models of gene regulation\, we also seek to  \n drive biological discovery through close collaborations with experimentalists  \n to study specific regulatory mechanisms.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10612.field_date.0.231
SUMMARY:Ankit Sharma: Welfare and Profit maximization with Procurement costs
DTSTAMP:20130618T164412Z
DTSTART:20121203T170000Z
DTEND:20121203T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/welfare-and-profit-maximization-procurement-costs
LOCATION:CS 2310
DESCRIPTION:Problem Statement A limited resource needs to be allocated amongst a set of  \n self-interested agents. Combinatorial Auctions\, that capture this situation\,  \n are a central problem in Algorithmic Mechanism Design: how to price and  \n allocate goods to buyers with complex preferences in order to maximize an  \n objective such as social welfare or revenue. Our Contributions The problem  \n has been well-studied in the case of limited supply (few copies of each  \n item)\, and unlimited supply. While in the case of limited supply\, there are  \n strong lower bounds known\, the case of unlimited supply may be too optimistic  \n for many settings. In this talk\, we look at how by relaxing the "strength" of  \n the limitation of available copies of an item\, we can beat the lower bounds  \n for the limited supply and get a smooth transition to the case of unlimited  \n supply. While if the limitation is "soft"\, we achieve a constant factor  \n approximation\, in case of "hard" limitation\, we achieve a logarithmic  \n approximation. Highlight of the Results In a work with Avrim Blum\, Anupam  \n Gupta and Yishay Mansour\, we initiate the study of resource allocation in a  \n more realistic modeling of limitation where each resource has an increasing  \n procurement cost. In this talk\, we describe a simple allocation scheme that  \n achieves a constant factor approximation to the optimal social welfare for  \n polynomial and logarithmic procurement cost curves. Furthermore\, for  \n arbitrary procurement cost curves\, we describe a scheme that achieves a  \n logarithmic approximation to optimal welfare. We consider arbitary  \n combinatorial valuation of the buyers and consider the two objective  \n functions of social welfare and revenue. The results are part of a work that  \n appeared at Foundations of Computer Science (FOCS)\, 2011. Joint work with  \n Avrim Blum\, Anupam Gupta and Yishay Mansour. Bio: Ankit Sharma is a graduate  \n student at Carnegie Mellon University. He is advised by Avrim Blum and Anupam  \n Gupta. His current research focuses on algorithms and algorithmic game  \n theory.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10613.field_date.0.232
SUMMARY:Let's help Johnny write robust applications: Let's help Johnny write robust  \n applications
DTSTAMP:20130618T164412Z
DTSTART:20121203T180000Z
DTEND:20121203T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/lets-help-johnny-write-robust-applications
LOCATION:4310
DESCRIPTION:Johnny wants to build robust applications---ones which will withstand attack.  \n But no matter how hard he tries\, the applications Johnny writes have security  \n holes. Experience shows that a large factor in Johnny's failure is the system  \n semantics on top of which his application runs. These semantics are difficult  \n for him to reason about\, have undefined and unspecified behaviors\, require  \n his application to use arcane security interfaces\, have hard-to-reason-about  \n corner cases\, and interact poorly. The result is a system with almost no  \n security guarantees. Ethos is an operating system being developed at the  \n University of Illinois at Chicago\, designed from the ground up to provide and  \n make it easier to reason about security properties. This talk will address  \n some particular interfaces in existing systems and show how Ethos improves on  \n the status quo. And a bio: Mike Petullo is a Ph.D. candidate at the  \n University of Illinois at Chicago. He is working with Dr. Jon Solworth on  \n Ethos\, an Operating System designed to provide application programmers with  \n semantics that make it easier to write secure code. Mike received a B.S. in  \n Computer Science from Drake University and a M.S. in Computer Science from  \n DePaul University. After graduating from UIC\, Mike will be serving as an  \n instructor in the Electrical Engineering and Computer Science Department at  \n the United States Military Academy.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10607.field_date.0.233
SUMMARY:Thomas Cox\, Professor: Brown Gold: The Smell of $$$ Optimizing  \n Economic/Environmental Sustainability via Manure Separation and  \n Bio-Feedstocks
DTSTAMP:20130618T164412Z
DTSTART:20121203T220000Z
DTEND:20121203T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/brown-gold-smell-optimizing-economicenvironmental-sustainability-manure-separation-and-bio-fee
LOCATION:Wisconsin Institute for...
DESCRIPTION:Biofeedstocks (dairy manure) from WI’ crucial dairy sector provide multiple  \n economic/environmental challenges and opportunities to “Grow the Green  \n Economy”. This biofeedstock shares fundamental (and relatively low cost)  \n bio-chemical processing – fermentation – with several classical WI Ag  \n products: beer\, cheese and kraut. Triple fermentation via 1) dairy feed  \n rations (corn/alfalfa silage\, haylage\, etc.)\, 2) the cow’s rumen\, and 3)  \n anaerobic digestion of manure provides an aggregated\, homogenized and  \n (bio-chemically) “pre-processed” (low cost) cellulosic feedstock – one  \n that largely avoids the food/fuel issues associated with many 1st and 2nd  \n generation bio-feedstocks. Extant and emerging separation technologies  \n provide opportunities to “fractionate” this cellulosic feedstock into  \n multiple value added products such as energy (methane gas\, ethanol)\,  \n intermediate chemical products (e.g.\, bio-butanol)\, mulch (peat moss  \n substitute)\, organic fertilizers and a variety of amino acids (protein  \n feedstocks for plastics). These value added opportunities enhance  \n sustainability\, both economic (increased sales/revenues and decreased costs  \n from on-farm substitution) AND environmental (improved carbon footprint/GHG  \n remediation\, reduced nutrient variability\, increased Precision Ag and reduced  \n environmental losses). This provides the fundamental sustainability basis for  \n the Economic/Environmental Win-Win and growing the local Green Economy (via  \n local manufacturing of the separation technology and growth in associated  \n bio-feedstock processing). UW Biomass R&D Initiative (BRDI) $7M Project:  \n Accelerated Renewable Energy • Tom Cox: UW Ag & Applied Economics • John  \n Norman: Emeritus\, UW Soil Science • Jim Leverich: On Farm Research  \n Coordinator\, UW-Extension WID-DOW: Optimizing Economic/Environmental  \n Sustainability • Michael Ferris & Hongbo Dong: UW Computer Science
END:VEVENT
BEGIN:VEVENT
UID:calendar.10504.field_date.0.234
SUMMARY:Paul E. Black: Software Assessment and the SAMATE Reference Dataset
DTSTAMP:20130618T164412Z
DTSTART:20121204T220000Z
DTEND:20121204T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/software-assessment-and-samate-reference-dataset
LOCATION:1240CS
DESCRIPTION:We take a few minutes to relate what NIST is\, opportunities to work with or  \n at NIST\, and the Software Assurance Metrics And Tool Evaluation (SAMATE)  \n project\, with efforts such as a workshop on Statistics\, the Universe of  \n Programs\, and Everything Relevant. Although assurance cannot be tested into  \n software\, software assessment can provide an additional component of  \n assurance.  Assessment can be broadly classified as static analysis or  \n dynamic analysis (testing)\, which have different strengths and which  \n complement each other. Planning for the fifth Static Analysis Tool Exposition  \n (SATE V) is in progress to deepen our understanding of code vulnerabilities\,  \n weaknesses\, and the abilities and limits of automated tools. We end with a  \n description of the SAMATE Reference Dataset (SRD)\, a public repository of  \n over 60\,000 reference programs.  We describe some of the test suites and  \n suggest ways they may help calibrate metrics and assessment techniques. Bio  \n Dr. Black has nearly 20 years of industrial experience in areas such as  \n developing software for IC design and verification\, assuring software  \n quality\, and managing business data processing.  He is a Computer Scientist  \n for the National Institute of Standards and Technology (NIST) in the Systems  \n and Software Division of the Information Technology Laboratory. Black earned  \n a Ph.D. at Brigham Young University in 1998.  He has taught classes at  \n Brigham Young University and Johns Hopkins University. He has published in  \n the areas of static analysis\, software testing\, formal methods\, software  \n verification\, quantum computing\, and computer forensics.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10618.field_date.0.235
SUMMARY:Sushmita Roy and Rob Atlas: AISEM 2012 - Meet the Roy Lab
DTSTAMP:20130618T164412Z
DTSTART:20121205T220000Z
DTEND:20121205T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/aisem-2012-meet-roy-lab
LOCATION:3310
DESCRIPTION:AISEM series provide a unique opportunity for both new and returning students  \n to interact with groups of the UW Artificial Intelligence community  \n (including machine learning\, bioinformatics\, computer vision etc.) and hear  \n about their research. Each seminar will feature a unique group. The seminar  \n includes several short talks given by students in that group and a social  \n time which the PI will be present. Refreshments are provided. In this  \n seminar\, we will introduce Prof. Sushmita Roy’s lab. The Roy lab is  \n interested in developing and applying machine learning algorithms to  \n understand the structure\, function and evolution of regulatory networks. Her  \n lab develops methods to integrate diverse regulatory genomics datasets to  \n identify such networks. Her lab is also interested in understanding how these  \n networks change across different biological contexts (e.g. different time  \n points\, different species)\, and how this affects the behavior of organisms.  \n Prof. Roy and one of her student Rob Atlas will present their works. Title:  \n Reconstructing condition-specific regulatory networks across long and short  \n time scales Speaker: Prof. Sushmita Roy Abstract: Changes in gene regulation  \n are hypothesized to play a major role in adaptation and evolution of  \n organism. In recent years\, functional genomics approaches have been used to  \n measure different aspects of the gene regulation machinery in single species  \n and extended to multiple species. These functional genomics datasets give us  \n the unique opportunity to develop more comprehensive models of gene  \n regulation. However\, doing so requires us to develop novel computational  \n tools that integrate such datasets within one species and across multiple  \n species. In this talk\, I will present computational methods to integrate  \n different types of datasets for the regulatory network of the model organism\,  \n Drosophila melanogaster. We show that data integration is key to improved  \n performance and increased coverage of the fly regulatory network. I will then  \n describe a multi-species analysis framework\, which comprises a (1) novel  \n multi-species clustering algorithm\, Arboretum that identifies modules in  \n species across large evolutionary distances\, and (2) a set of metrics to  \n examine patterns of conservation and divergence in these modules\, as well as  \n the factors that drive divergence. We applied our approach to expression  \n profiles measured in 8 species of Ascomycota fungi under glucose depletion  \n and 8 species under heat shock. In both responses\, the transcriptomes are  \n captured by five conserved expression modules\, however\, the degree of gene  \n content conservation in the module was substantially lower in heat shock than  \n glucose depletion\, suggesting a stronger conservation of the latter response.  \n Our approaches for integrating different types of regulatory datasets\, within  \n one organism and across multiple organisms\, can lead us to systematically  \n understand the structure\, function\, and evolution of regulatory networks.  \n Title: A graph-based comparative analysis of three-dimensional organization  \n of chromosomes in yeast and mammals Speaker: Rob Atlas Abstract: Genome-wide  \n maps of chromosomal interactions are becoming increasingly common.  \n Computational tools to analyze these maps and\, more importantly\, to compare  \n them across contexts are scarce. In this talk I will describe a novel  \n graph-based clustering approach for detecting sets of interacting genomic  \n loci and comparing them across different tissues and organisms. Our hybrid  \n approach can be applied analogously to 3C data from simple eukaryotes as well  \n as higher eukaryotes such as mouse and humans. We use a penalized cluster  \n quality criteria to determine the number of clusters and develop numerous  \n statistics to systematically examine properties of chromosomal organization.  \n Application of our approach identifies several properties: (a) the proportion  \n of inter-chromosomal interactions is much higher in yeast compared to  \n mammalian species\, (b) in addition to replication forks\, distally interacting  \n regions in yeast also exhibit a tendency to be co-regulated (based on gene  \n expression\, ChIP- or genetic knockout-based targets of transcription  \n factors)\, (c) most chromosomal organization is similar between two mammalian  \n tissues\, but there are regions that exhibit a tissue-specific interaction  \n pattern\, (d) comparison of interaction maps between human and mouse identify  \n extensive conservation of clusters of regions\, but we did not find similar  \n conservation between yeast and mammals. Our graph-based clustering method  \n facilitates a systematic comparison of chromosomal region interaction maps  \n across tissues and species\, enabling us to corroborate known findings and  \n identify novel aspects of such interaction maps.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10621.field_date.0.236
SUMMARY:William Freeman: Seeing Things That are Hard to See
DTSTAMP:20130618T164412Z
DTSTART:20121206T220000Z
DTEND:20121206T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/seeing-things-are-hard-see
LOCATION:1240
DESCRIPTION:Abstract:I will describe two recent projects that both seek to reveal things  \n in images or videos that are otherwise difficult to see. Video magnification  \n amplifies small motion or color changes in videos\, allowing a real-time  \n "microscope" to view otherwise invisible changes. (Joint work with Michael  \n Rubinstein\, Hao-Yu Wu\, Neal Wadhwa\, Eugene Shih\, John Guttag\, and Fredo  \n Durand). The second project\, accidental camera\, reveals information about  \n structures outside the frame of a photograph or a video through the  \n relatively common formation of accidental cameras. (Joint work with Antonio  \n Torralba). Bio: William T. Freeman is Professor of Electrical Engineering and  \n Computer Science at the Computer Science and Artificial Intelligence  \n Laboratory (CSAIL) at MIT\, joining the faculty in 2001. His current research  \n interests include machine learning applied to computer vision\, Bayesian  \n models of visual perception\, and computational photography. He received  \n outstanding paper awards at computer vision or machine learning conferences  \n in 1997\, 2006\, 2009 and 2012. Previous research topics include steerable  \n filters and pyramids\, the generic viewpoint assumption\, color constancy\,  \n computer vision for computer games\, and bilinear models for separating style  \n and content. He is active in the program or organizing committees of computer  \n vision\, graphics\, and machine learning conferences. He was the program  \n co-chair for ICCV 2005\, and is program co-chair for CVPR 2013.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10614.field_date.0.237
SUMMARY:Hamid Reza Ghasemi: Architecture and Circuit Techniques for Power-efficient  \n Multicore Processors
DTSTAMP:20130618T164412Z
DTSTART:20121207T150000Z
DTEND:20121207T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/architecture-and-circuit-techniques-power-efficient-multicore-processors
LOCATION:CS 3310
DESCRIPTION:Committee: Nam Sung Kim (advisor)\, Mark D. Hill\, Mikko H. Lipasti\, Karu  \n Sankaralingam\, Mike Schulte
END:VEVENT
BEGIN:VEVENT
UID:calendar.10620.field_date.0.238
SUMMARY:Haris Volos: Revamping the System Interface to Storage-Class Memory
DTSTAMP:20130618T164412Z
DTSTART:20121210T160000Z
DTEND:20121210T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/revamping-system-interface-storage-class-memory
LOCATION:2310 Computer Sciences
DESCRIPTION:Committee: Michael Swift (advisor)\; Andrea Arpaci-Dusseau\, Shan Lu\, Jignesh  \n Patel\, David Wood
END:VEVENT
BEGIN:VEVENT
UID:calendar.10626.field_date.0.239
SUMMARY:Analysis Of Software Artifacts – CS706 Project Presentations\, Part I
DTSTAMP:20130618T164412Z
DTSTART:20121210T203000Z
DTEND:20121210T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/analysis-software-artifacts-%E2%80%93-cs706-project-presentations-part-i
LOCATION:1240 CS
DESCRIPTION:You are cordially invited to attend the class project presentations for  \n CS706: Analysis of Software Artifacts. Come learn about a variety of  \n innovative approaches to helping humans build software that works. Attend as  \n many or as few talks as you wish. 2:30pm – 2:45pm: Benefits of Aspect  \n Oriented Programming on Code Injection 2:45pm – 3:00pm: XPose: Finding Flow  \n Dependences Across the Operating System 3:00pm – 3:15pm: Mining Software  \n Repositories for Accurate Authorship 3:15pm – 3:30pm: Vizmake: Debugging  \n Makefiles via Visualization Visit the complete CS706 Project Presentation  \n Schedule for more information\, including talk abstracts. See also additional  \n CS706 presentation sessions on Wednesday\, December 12 and Friday\, December  \n 14.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10611.field_date.0.240
SUMMARY:Shuchi Chawla\, Assistant Professor : Approximation in Mechanism Design
DTSTAMP:20130618T164412Z
DTSTART:20121210T220000Z
DTEND:20121210T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/approximation-mechanism-design
LOCATION:Wisconsin Institute for...
DESCRIPTION:Mechanism design is the subfield of game theory that deals with economic  \n systems from the point of view of an optimizing designer. It involves  \n optimization problems where the input to the problem is owned by  \n self-interested parties (a.k.a. agents). The designer would like to come up  \n with a protocol whereby the individual optimizations of the agents lead in  \n equilibrium to optimization of the global objective. In this talk we will  \n discuss some classical problems in mechanism design and the role of  \n approximation in solving them. No knowledge of game theory is necessary.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10631.field_date.0.241
SUMMARY:Gary Don Pack: Semiparametric Geometric Methods for Extracting and Modeling  \n White Matter Volumetric Structures of the Brain
DTSTAMP:20130618T164412Z
DTSTART:20121211T200000Z
DTEND:20121211T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/semiparametric-geometric-methods-extracting-and-modeling-white-matter-volumetric-structures-br
LOCATION:3310 Computer Scinces
DESCRIPTION:Committee: Charles Dyer (advisor)\, Andrew Alexander\, M. Elizabeth Meyerand\,  \n Moo Chung\, Vikas Singh
END:VEVENT
BEGIN:VEVENT
UID:calendar.10622.field_date.0.242
SUMMARY:CSEdWeek Scratch Tutorial
DTSTAMP:20130618T164412Z
DTSTART:20121211T220000Z
DTEND:20121211T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/csedweek-scratch-tutorial
LOCATION:Mac Lab\, CS 1st Floor
DESCRIPTION:December 9-15 is CS Education Week (http://www.csedweek.org/). In this  \n spirit\, we're holding a tutorial about Scratch. Scratch is a programming  \n language that makes it easy to create interactive stories\, animations\, games\,  \n music\, and art. Scratch is easy to learn and surprisingly powerful. Scratch  \n is frequently used for Computer Science outreach\, and is currently used in  \n our own CS 202 and CS 402. Come have fun using Scratch and learn about using  \n it as a computer science teaching tool!
END:VEVENT
BEGIN:VEVENT
UID:calendar.10633.field_date.0.243
SUMMARY:Chris HInrichs: Multi-Modality Inference Methods for Neuroimaging with  \n Applications to Alzheimer's Disease Research
DTSTAMP:20130618T164412Z
DTSTART:20121212T200000Z
DTEND:20121212T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/multi-modality-inference-methods-neuroimaging-applications-alzheimers-disease-research
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: Vikas Singh (advisor)\, Sterling Johnson\, Moo K. Chung\, Grace  \n Wabha\, Charles Dyer
END:VEVENT
BEGIN:VEVENT
UID:calendar.10627.field_date.0.244
SUMMARY:Analysis Of Software Artifacts – CS706 Project Presentations\, Part II
DTSTAMP:20130618T164412Z
DTSTART:20121212T203000Z
DTEND:20121212T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/analysis-software-artifacts-%E2%80%93-cs706-project-presentations-part-ii
LOCATION:1240 CS
DESCRIPTION:Analysis Of Software Artifacts – CS706 Project Presentations\, Part I 1240  \n CS University of Wisconsin–MadisonYou are cordially invited to attend the  \n class project presentations for CS706: Analysis of Software Artifacts. Come  \n learn about a variety of innovative approaches to helping humans build  \n software that works. Attend as many or as few talks as you wish. 2:30pm –  \n 2:45pm: Notification of Propagating Source Code Changes Affecting Use Cases  \n 2:45pm – 3:00pm: Random hunting for stale bugs in the Linux kernel 3:00pm  \n – 3:15pm: Entropy as a Measure of Software Size 3:15pm – 3:30pm:  \n Identifying Performance Bugs Inside Loops Visit the complete CS706 Project  \n Presentation Schedule for more information\, including talk abstracts. See  \n also additional CS706 presentation sessions on Monday\, December 10 and  \n Friday\, December 14.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10632.field_date.0.245
SUMMARY:Xiaoyong Chai: Building Structured Web Portals:Research Challenges and  \n General Platforms
DTSTAMP:20130618T164412Z
DTSTART:20121213T160000Z
DTEND:20121213T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/building-structured-web-portalsresearch-challenges-and-general-platforms
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: AnHai Doan (advisor)\, Jeffrey Naughton\, Jignesh Patel\, Jude  \n Shavlik\, Mark Craven
END:VEVENT
BEGIN:VEVENT
UID:calendar.10650.field_date.0.246
SUMMARY:Jongwon Yoon: Resource Management Schemes in OFDMA Networks
DTSTAMP:20130618T164412Z
DTSTART:20121213T160000Z
DTEND:20121213T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/resource-management-schemes-ofdma-networks
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: Suman Banarjee (advisor)\, Aditya Akella and Parmesh Ramanathan
END:VEVENT
BEGIN:VEVENT
UID:calendar.10610.field_date.0.247
SUMMARY:Somayeh Sardashti: Energy-Optimized Memory Hierarchies in Multi-Core  \n Processors
DTSTAMP:20130618T164412Z
DTSTART:20121214T163000Z
DTEND:20121214T183000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/energy-optimized-memory-hierarchies-multi-core-processors
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: David Wood (advisor)\, Mark Hill\, Guri Sohi\, Nam Sung Kim\, Remzi  \n Arpaci-Dusseau
END:VEVENT
BEGIN:VEVENT
UID:calendar.10576.field_date.0.248
SUMMARY:Hakim Weatherspoon: Plug into the Supercloud
DTSTAMP:20130618T164412Z
DTSTART:20121214T190000Z
DTEND:20121214T200000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/plug-supercloud
LOCATION:CS 1240
DESCRIPTION:Abstract: Cloud computing is often compared to the power utility model as  \n part of a trend towards the commoditization of computing resources. However\,  \n today’s cloud providers do not simply supply raw computing resources as a  \n commodity\, but also act as distributors\, dictating cloud services that are  \n not compatible across providers. In this talk\, I will discuss a new cloud  \n service distribution layer\, called a Supercloud\, that is completely decoupled  \n from the cloud provider. A Supercloud give its users the illusion of their  \n own homogenized private cloud (albeit\, layered on top of one or more  \n third-party providers). Under the hood\, the Supercloud can include different  \n hypervisors\, hardware architectures\, storage subsystems\, and connectivity  \n fabrics. Leveraging a nested paravirtualization layer called the Xen-Blanket\,  \n the Supercloud maintains the control necessary to implement hypervisor-level  \n services and management. Using the Xen-Blanket to transform various cloud  \n provider services into a unified offering\, we have deployed a Supercloud  \n across Amazon’s Elastic Compute Cloud (EC2)\, IBM\, and Cornell University\,  \n and performed live VM migration between the different sites. Furthermore\,  \n Superclouds create opportunities to exploit resource management techniques  \n that providers do not expose\, like resource oversubscription\, and ultimately  \n can reduce costs for users. Speaker Bio: Hakim Weatherspoon is an assistant  \n professor in the Department of Computer Science at Cornell University. His  \n research interests cover various aspects of fault-tolerance\, reliability\,  \n security\, and performance of large Internet-scale systems such as cloud  \n computing and distributed systems. Professor Weatherspoon received his Ph.D.  \n from University of California at Berkeley and B.S. from University of  \n Washington. He is an Alfred P. Sloan Fellow and recipient of an NSF CAREER  \n award\, DARPA Computer Science Study Panel (CSSP)\, IBM Faculty Award\, the  \n NetApp Faculty Fellowship\, Intel Early Career Faculty Honor\, and the Future  \n Internet Architecture award from the National Science Foundation (NSF).
END:VEVENT
BEGIN:VEVENT
UID:calendar.10628.field_date.0.249
SUMMARY:Analysis Of Software Artifacts – CS706 Project Presentations\, Part III
DTSTAMP:20130618T164412Z
DTSTART:20121214T203000Z
DTEND:20121214T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/analysis-software-artifacts-%E2%80%93-cs706-project-presentations-part-iii
LOCATION:1240 CS
DESCRIPTION:You are cordially invited to attend the class project presentations for  \n CS706: Analysis of Software Artifacts. Come learn about a variety of  \n innovative approaches to helping humans build software that works. Attend as  \n many or as few talks as you wish. 2:30pm – 2:45pm: GPGPU Centrifuge: Using  \n Dynamic Profiling to Reindex Threads 2:45pm – 3:00pm: Exploring the state  \n space of mouse event interleavings in Javascript applications 3:00pm –  \n 3:15pm: Coding Style Extractors 3:15pm – 3:30pm: Learning About Open Source  \n Software from Github’s Geographical Data Visit the complete CS706 Project  \n Presentation Schedule for more information\, including talk abstracts. See  \n also additional CS706 presentation sessions on Monday\, December 10 and  \n Wednesday\, December 12.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10635.field_date.0.250
SUMMARY:CS Graduation Reception
DTSTAMP:20130618T164412Z
DTSTART:20121216T180000Z
DTEND:20121216T200000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/cs-graduation-reception
LOCATION:University Club 803...
DESCRIPTION:CS Department Graduation Reception for Ph.D.\, Masters and Undergraduate  \n Students who are graduating this semester. By RSVP required by Dec. 11.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10634.field_date.0.251
SUMMARY:Ning Zhang: Towards Cost-Effective Resources Provisioning for DBMS and  \n Storage Clouds
DTSTAMP:20130618T164412Z
DTSTART:20121217T160000Z
DTEND:20121217T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/towards-cost-effective-resources-provisioning-dbms-and-storage-clouds
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: Jignesh Patel (advisor)\, Jeffrey Naughton\, Christopher Re\, Hakan  \n Hacigumus\, Rafael Lazimy
END:VEVENT
BEGIN:VEVENT
UID:calendar.10683.field_date.0.252
SUMMARY:CS/Psych-770 Human-Computer Interaction Poster Session
DTSTAMP:20130618T164412Z
DTSTART:20121218T210000Z
DTEND:20121218T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/cspsych-770-human-computer-interaction-poster-session
LOCATION:CS Lobby\, West Entrance...
DESCRIPTION:We invite graduate students and faculty to the CS/Psych-770 Human-Computer  \n Interaction Poster Session to hear about 11 very interesting research  \n projects completed by interdisciplinary groups of students. The Poster  \n Session will be on the West Entrance Ramp of the Computer Sciences Building  \n at 3-5 pm on Tuesday\, December 18\, 2012. There will be chocolate and cookies.  \n Below are abstract for the projects.The effects of Robot Orientation on  \n Robot-mediated Group ConversationsTsu-Lun Huang & Guru  \n SubramaniAbstract: Although telepresence robot mediated communication is  \n sparsely used at present\, it is expected to be pervasive in the future. The  \n goal of this study is to investigate the effect of ability to control the  \n orientation of a telepresence robot has on group discussions. We tested (1)  \n how telepresence robot mediated communication can be improved in terms of  \n conversation quality and involvement\, and (2) how the group perceived the  \n utilization of a telepresence robot. We compared conversations mediated  \n through traditional video conferences with conversations mediated through  \n telepresence robot. Two experiments were conducted to examine the perception  \n of both members on screen and physically in the group. A higher level of  \n involvement was found when the robot can move its orientation. It is  \n suggested that the utilization of telepresence robot with orienting features  \n can be beneficial in group scenarios. More research is warranted with regards  \n to the add-on feature of a telepresence robot.The Effects of Agents on  \n LyingMonica Thompson & Le YuAbstract: Studies have shown that people have a  \n tendency to lie on a regular basis. However\, how does this behavior change  \n when the entity being lied to happens to be a computer? Furthermore\, will the  \n tendency to lie change if the content of the lie is changed? Participants  \n were asked to evaluate their willingness to lie to a text-based computer  \n agent in ten different scenarios. Five of these scenarios dealt with  \n situations in which the participants were potentially motivated to lie about  \n a personal fact\, such as weight\, income\, and schedule\, while the remaining  \n five involved lying about a personal opinion\, which included musical taste\,  \n politics\, and religion. These participants were then compared to a control  \n group that evaluated the same scenarios\, except they evaluated their  \n willingness to lie to a person over text-mediated communication.Accommodating  \n Learning Styles: The Effect of Personalization on Learning PerformanceMaggie  \n Simon & Zexian ZengAbstract: Does personalization of a learning task  \n according to a user’s learning style affect the learning performance of  \n students? The literature shows students’ learning styles can affect overall  \n performance in a Computer Science class. We focus on two main learning  \n styles: visual and verbal. Twenty-five undergraduate students in CS 302  \n Introduction to Programming completed an online study\, which included a  \n learning styles inventory\, pre-test\, learning task\, post-test\, and a  \n questionnaire about the study. Visual learners received a more visually  \n oriented learning task while verbal learners received a more verbally  \n oriented learning task. Those in the control group randomly received either  \n the visual or verbal task. The primary measures include the percentage of  \n questions answered correctly on the learning task and subjective self-report  \n measures from the questionnaire. With this information\, we evaluate the  \n effect of personalization based on learning style.The Effects of Background  \n Environment and Gender on Social Presence and Trust During  \n TelecollaborationRebecca Perkins & Alex PeerAbstract: "Social presence  \n refers simply to the feeling of being together with other parties engaged in  \n a communication activity.” (Isgro et al. 2004\, p.288) and is desirable  \n during telecollaboration as it has been associated with increased trust and  \n self-disclosure (Gunawardena\, 2001\, p.115). During telecollaboration\, images  \n of participants may highlight differences in user environments. This  \n experiment examined how varying levels of environment similarity affect  \n users’ sense of social presence and trust during telecollaboration and  \n whether those effects vary based on participant gender. Participants were  \n asked to complete a self-disclosure interview followed by a survey designed  \n to measure social presence and trust. Subjective measures were the survey  \n results and objective measures were answer duration and number of items  \n shared with the interviewer. No significant effect of environment on social  \n presence or trust was found and no significant interaction effect was found  \n between environment and gender.Effects of Computer-Mediated Communication  \n Software Reliability on Trust Between CollaboratorsSarah Gilliland & Taylor  \n PattersonAbstract: It is becoming more frequent in industry and academia for  \n collaboration to occur across considerable spacial and temporal separations.  \n Previous research has shown that in the absence of face-to-face contact\, the  \n establishment and maintenance of trust between two or more people is  \n observably more difficult. In addition\, many software applications used for  \n distance collaboration suffer from poor reliability. The goal of this study  \n was to determine whether faulty or unreliable software affects the trust  \n between two people collaborating on a project. We performed a  \n between-participants study in which two people\, a participant and a  \n confederate\, played a social dilemma game with only an instant messaging  \n system to communicate with each other. In the control group\, the  \n communication software worked flawlessly while the evaluation group was  \n subjected to frequent connection issues and failed message delivery.  \n Objective and subjective analyses were completed in order to evaluate  \n participants' trust in their collaboration partner.The effects of Cognitive  \n Load and Error Correcting Mode on User Preference for Voice Recognition  \n AgentsShiyu Luo & Theodora HinkleAbstract: Voice recognition technology is  \n growing rapidly nowadays\, it gives rise to growing groups of users. However\,  \n this technology is far from being perfect\, thus error correction is  \n necessary. We are interested in investigating people's attitudes towards this  \n technology and their preferences for different error correction modes under  \n different cognitive loads. A 2x2 factor design is conducted. Participants are  \n asked to dictate different script to a voice recognition agent while they are  \n playing a game\, which simulates cognitive load in reality. Our study shows  \n that people generally have a grown positive attitude in terms of accuracy and  \n satisfaction towards voice recognition after completing our experiments.  \n Also\, our analysis shows that people lean toward the agent that offers  \n suggestions when an error occurs\, as oppose to the one that simple asks for  \n repetition.Effects of using computer agents vs. human on participant’s  \n information exposure in health interviewsXishuo Liu & Guangyu  \n LiuAbstract: Agent technologies have been widely considered in health care  \n area. However there is little scientific research that considers information  \n disclosures by people when interviewed by computer agents. In this work\, we  \n study information disclosures from male and female when interviewed by human  \n versus computer agent. We recruit participants from the University of  \n Wisconsin-Madison and Amazon Mechanical Turks. In our experiments\,  \n participants are interviewed on health care questions by a human interviewer  \n or a computer agent. Participants answer interview questions by typing their  \n answers into text boxes. From our data analyses\, we show that the people are  \n more likely to disclose information to human interviewer than computer agent.  \n We also see that female participants disclose more information than male  \n participants on normal health care questions but less likely to answer much  \n on health related questions.Leveraging Users Preference over Cloud  \n Collaborative SoftwareXiayuan Huang & Xiujun LiAbstract: The advent of cloud  \n collaborative software or platform has brought convenience to people for file  \n storage\, sharing and document collaboration\, and it is being more and more  \n popular among people. However\, not all their functionalities and features are  \n inspiring. We conduct an explorative study to leverage users preference and  \n trust over some certain features (system encryption\, user controls and  \n service payment) to get an idea about people’s tradeoff when they are using  \n this kind of software\, to explore the correlationship among these  \n features. Effect of guidance from an agent for online foreign language  \n studyDaniel Crowell & Jignwei LiAbstract: The emergence of new interactive  \n technologies could bring an opportunity to improve traditional learning  \n methods and be the bright future for studying a foreign language. Our goal of  \n the study is to see how much effect an agent or an avatar can have to assist  \n people learning and memorizing foreign language words. The foreign language  \n that we chose to conduct the research is Dutch. We used text and audio  \n methods to make comparisons with the agent guidance. In the text option\,  \n participants only saw the Dutch word with a corresponding English word. The  \n audio setting involved the participants seeing the word and the corresponded  \n English word as well as hearing the word pronunciation and definition. For  \n the last option\, the participants saw agents presenting the meaning of the  \n word. The 36 Dutch words and their presentation type (text\, audio\, or agent)  \n were both randomly shuffled. At the end of the experiment\, a memory test  \n aimed to evaluate how well the participants memorized the words. We want to  \n explore whether the presentations given by an agent would help the  \n participant on memory retention. We are also curious if the words that were  \n shown earlier would end up with higher correctness in the test than the ones  \n that were shown later. The participants were gathered through Amazon  \n Mechanical Turk\, and agents were made through thevoki.com\, a website for  \n creating virtual speaking avatars.Effects of information representation on  \n recall in learning environment: a study of text\, audio\, multimedia\, and  \n human-like embodied agentYuqi He & Nai-Wen YuAbstract: A multimedia  \n instructional message is a presentation consisting of words and pictures that  \n is designed to foster meaningful learning. There is increasing interest in  \n use of human embodied agents\, also known as avatars\, in the learning  \n environment\, specifically\, in e-learning environment. Human-embodied agents  \n have effects on human cognition and potentially increase students’  \n enjoyment of the learning experiences and students’ motivation. Our team  \n notices this future trend\; therefore\, the goal of this study is to compare  \n how different learning strategies— text\, audio\, multimedia and human like  \n embodied agent—would affect students’ recall ability and examine whether  \n human embodied agents would be superior than other learning strategies on  \n recall. The topic we choose to test is a rare disease: Golloway-Mowat  \n Syndrome. Thirty two native speakers were recruited in our study. We  \n conducted a between-participant experimental design to evaluate which method  \n has a stronger impact on recall performances.The Effects of False Accusation  \n in Automated Adjustment Notices on Procedural Fairness Perceptions of the  \n IRS\, Social Representations of Taxes\, and Future ComplianceCass  \n HaussermanAbstract: This study examines how an error by the IRS accusing  \n taxpayers of owing additional money affects perceptions of procedural  \n fairness\, social representations of taxes\, and ultimately future compliance.  \n In a 2 x 2 fully factorial design\, I manipulate whether the IRS system making  \n the error is automated or not\, as well as the party at fault for the error  \n – either the IRS or the taxpayer’s employer. Twenty participants complete  \n this study in which they are randomly assigned to conditions and receive  \n hypothetical information about a scenario. Participants make compliance  \n decisions and provide information about their feelings\, as well as various  \n demographic data and other potential controls. Both independent variables are  \n marginally significant in predicting future compliance\, but not necessarily  \n in the expected directions. Automation is marginally significant in  \n predicting social representations of taxes\, and both automation and source of  \n error are significant in predicting procedural fairness. Specifically\,  \n procedural fairness is higher when the system is automated and lower when the  \n IRS made the error. In addition to these preliminary findings\, results and  \n feedback from this study provide insight into necessary instrument  \n improvements. 
END:VEVENT
BEGIN:VEVENT
UID:calendar.11363.field_date.0.253
SUMMARY:Wolfgang Gaggl\, PhD: Functional MRI\, Connectivity\, and DTI in Clinical  \n Practice
DTSTAMP:20130618T164412Z
DTSTART:20130122T220000Z
DTEND:20130122T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/functional-mri-connectivity-and-dti-clinical-practice
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) Seminar: Wolfgang  \n Gaggl\, PhD\, received his PhD in Biophysics (MRI) at the Medical College of  \n Wisconsin\, Milwaukee\, in 2012. He also holds a Master in Electrical  \n Engineering from the University of Technology in Graz\, Austria\, with focus on  \n digital signal processing. After receiving his master degree in 2001\, he  \n joined the Department of Otolaryngology and Communication Sciences at the  \n Medical College of Wisconsin\, where he worked on EEG and functional MRI  \n (FMRI) of the auditory system and collaborated on studies on cochlear  \n implants. In 2004 he moved to the Radiology Department as a senior research  \n engineer for MRI where he specialized in Diffusion Tensor Imaging (DTI). He  \n created a seamless multimodal (CT\, MRI\, FMRI\, DTI) clinical imaging workflow  \n from the scanner into the operating room and initiated a close collaboration  \n with the Department of Neurosurgery to integrate these complementary imaging  \n modalities into the surgical navigation system. With his effort the use of  \n DTI and FMRI in the surgical resection of brain tumors has increased  \n approximately 10 fold at the Medical College over the past 6 years\, which led  \n to a significant reduction of post-surgical deficits. For his doctoral thesis  \n Dr. Gaggl developed a high-resolution DTI method that allows clinical DTI  \n acquisitions in human subjects at 0.4mm in-plane resolution\, thereby  \n considerably improving the capabilities of clinical DTI in human brain and  \n spinal cord.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11388.field_date.0.254
SUMMARY:Caroline Uhler: Geometry of Gaussian Graphical Models
DTSTAMP:20130618T164412Z
DTSTART:20130123T220000Z
DTEND:20130123T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/geometry-gaussian-graphical-models
LOCATION:Bardeen Med Lab Building...
DESCRIPTION:One of the main problems related to Gaussian graphical models is learning\,  \n i.e. the estimation of model parameters and structure. In this talk\, we apply  \n algebraic geometry to analyze two widely used methods for learning\, namely  \n the maximum likelihood approach for parameter estimation and the PC-algorithm  \n for structure estimation. First\, we give an algebraic criterion to find exact  \n lower bounds on the number of observations needed for the existence of the  \n maximum likelihood estimator in undirected Gaussian graphical models. This is  \n particularly important for applications such as gene association networks\,  \n where the number of random variables is much larger than the number of  \n observations. We also find a first instance of a graph for which only  \n treewidth-many observations are needed\; an encouraging result. We then turn  \n to structure estimation in directed Gaussian graphical models and give a  \n rather discouraging result. We analyze the so-called "strong-faithfulness  \n condition"\, one of the main assumptions of the PC-algorithm\, and show that  \n this assumption is in fact extremely restrictive\, implying fundamental  \n limitations for the PC-algorithm and other algorithms based on partial  \n correlations.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11381.field_date.0.255
SUMMARY:Khai Tran: Realizing Parallelism in OLTP Workloads
DTSTAMP:20130618T164412Z
DTSTART:20130124T150000Z
DTEND:20130124T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/realizing-parallelism-oltp-workloads
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: Jeffrey Naughton (advisor)\, Jignesh Patel\, AnHai Doan\, Christopher  \n Re\, Jon Eckhardt
END:VEVENT
BEGIN:VEVENT
UID:calendar.10689.field_date.0.256
SUMMARY:Kevin Murphy: TBA
DTSTAMP:20130618T164412Z
DTSTART:20130124T220000Z
DTEND:20130124T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/tba-1
LOCATION:1240
DESCRIPTION:TBA
END:VEVENT
BEGIN:VEVENT
UID:calendar.11382.field_date.0.257
SUMMARY:Yueh-Hsuan Chiang: Matching Temporal Records
DTSTAMP:20130618T164412Z
DTSTART:20130125T170000Z
DTEND:20130125T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/matching-temporal-records
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: Jeffrey Naughton (advisor)\, AnHai Doan\, Christopher Re
END:VEVENT
BEGIN:VEVENT
UID:calendar.11444.field_date.0.258
SUMMARY:Pascale Carayon: Designing Health Information Technology for Clinicians and  \n Patients: A Human Factors Engineering Viewpoint
DTSTAMP:20130618T164412Z
DTSTART:20130129T220000Z
DTEND:20130129T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/designing-health-information-technology-clinicians-and-patients-human-factors-engineering-view
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) Seminar: Pascale  \n Carayon is Procter & Gamble Bascom Professor in Total Quality in the  \n Department of Industrial and Systems Engineering and the Director of the  \n Center for Quality and Productivity Improvement (CQPI) at the University of  \n Wisconsin-Madison. She received her Engineer diploma from the Ecole Centrale  \n de Paris\, France\, in 1984 and her Ph.D. in Industrial Engineering from the  \n University of Wisconsin-Madison in 1988. Her research areas include systems  \n engineering\, human factors and ergonomics\, sociotechnical engineering and  \n occupational health and safety. She is a scientific editor for Applied  \n Ergonomics and a member of the editorial board of the Journal of Patient  \n Safety. She is the chair of the technical committee on Organizational Design  \n And Management of the International Ergonomics Association (IEA)\, and is a  \n member of the executive committee of the IEA\, in charge of the Ergonomics In  \n Quality Design (EQUID) program and chair of the Science\, Technology and  \n Practice committee.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11420.field_date.0.259
SUMMARY:Joo-young Hwang: Flash Friendly File System (F2FS) for Mobile Flash Storages
DTSTAMP:20130618T164412Z
DTSTART:20130130T180000Z
DTEND:20130130T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/flash-friendly-file-system-f2fs-mobile-flash-storages
LOCATION:3310 CS
DESCRIPTION:Abstract: Recent mobile devices adopt various flash storage devices as a  \n primary storage. File system support for those devices is a must for flash  \n device performance and lifespan. I will present a new file system\, called  \n F2FS\, designed for mobile flash storage. F2FS is designed considering the  \n characteristics of the underlying flash storage which has a flash translation  \n layer (FTL). F2FS outperforms EXT4\, which is a popular file system for  \n Android phones\, in most benchmarks. I will describe the motivation\, design\,  \n and implementation of the file system\, then show a detailed performance  \n comparison with EXT4. Target audiences are those who are interested in file  \n system support for flash storages such as eMMC and SSD. Kernel and file  \n system expertise helps but is not mandatory to listen to this talk. Bio:  \n Joo-Young Hwang received Ph.D from KAIST in 2003 and has been working for  \n Samsung Electronics Co.\, Ltd. since then. His research interests include  \n storage\, file system\, virtualization\, and the Linux kernel. He developed RFS  \n which is Samsung's commercial file system for flash memory and a Xen  \n hypervisor for the ARM CPU. Recently he is working on developing storage  \n stack for flash storages such as SSD and eMMC cards.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11390.field_date.0.260
SUMMARY:Guanghui (George) Lan: Reduced-order Methods for Big-Data Challenges in  \n Stochastic and Nonlinear Optimization
DTSTAMP:20130618T164412Z
DTSTART:20130131T180000Z
DTEND:20130131T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/reduced-order-methods-big-data-challenges-stochastic-and-nonlinear-optimization
LOCATION:Mechanical Engineering...
DESCRIPTION:This talk focuses on the design and analysis of reduced-order methods to  \n tackle the big-data challenges in optimization. The last several years have  \n seen an unprecedented growth in the amount of available data. While  \n nonlinear\, especially convex programming models are important to extract  \n useful knowledge from raw data\, high problem dimensionality\, large data  \n volumes\, and inherent uncertainty present significant challenges to the  \n design of optimization algorithms. Aiming to attack some of these challenges\,  \n we introduce: i) a new class of stochastic approximation algorithms that can  \n yield the optimal rate of convergence for solving different stochastic  \n optimization problems. Some of these optimal rates were obtained for the  \n first time in the literature\; and ii) a new class of deterministic  \n first-order algorithms that can converge optimally\, require no structural  \n information and do not rely on line search\, based on level methods. To the  \n best of our knowledge\, no such uniformly optimal first-order methods have  \n been studied before in the literature. Applications of these  \n stochastic/deterministic algorithms will be studied. We will also briefly  \n discuss some other related work and possible future research directions. BIO:  \n Guanghui (George) Lan obtained his Ph.D. degree in Industrial and Systems  \n Engineering from Georgia Tech in 2009. He joined the Department of Industrial  \n and Systems Engineering at the University of Florida as an assistant  \n professor thereafter. His main research interests lie in the theory and  \n algorithms for stochastic optimization\, nonlinear programming\,  \n simulation-based optimization\, and their applications in various fields\, such  \n as large-scale data analysis and homeland security. His research has been  \n supported by the National Science Foundation and Office of Naval Research.  \n The academic honors that he received include the INFORMS Computing Society  \n Student Paper Competition First Place (2008)\, INFORMS George Nicholson Paper  \n Competition Second Place (2008)\, Mathematical Optimization Society Tucker  \n Prize Finalist (2012)\, INFORMS Junior Faculty Interest Group (JFIG) Paper  \n Competition First Place (2012) and the recent National Science Foundation  \n CAREER Award Winner (2013).
END:VEVENT
BEGIN:VEVENT
UID:calendar.11386.field_date.0.261
SUMMARY:Dick Sites: How Fast is My Disk?
DTSTAMP:20130618T164412Z
DTSTART:20130131T220000Z
DTEND:20130131T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/how-fast-my-disk
LOCATION:1240CS
DESCRIPTION:Modern hard drives contain read and write buffers to smooth the flow of data  \n to and from the disk surfaces. While these are beneficial\, the write buffer  \n can also introduce unexpected delays. This directly affects the  \n 99th-percentile latency of latency-sensitive reads when reads and writes are  \n intermixed. We discuss simple ways to measure from a C program what is  \n happening at the heads of a disk\, and a non-obvious technique for speeding up  \n reads without slowing down writes. The same simple programs can measure SSD  \n behavior. The same insights apply to understanding possible excessive  \n web-access latency in your cable modem. The subtext of the talk is that there  \n is no need to accept unknown hardware or software as "black boxes" that  \n cannot be further understood. Well-designed simple experiments can lead to  \n substantial insight. Some thought required.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11490.field_date.0.262
SUMMARY:Paul Kantor: Sensor Synergy
DTSTAMP:20130618T164412Z
DTSTART:20130204T193000Z
DTEND:20130204T203000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/sensor-synergy
LOCATION:Rm 1240
DESCRIPTION:Sensors\, or tests\, are central in our analysis of the state of our  \n surroundings. We examine the possible relations among two sensors which are  \n both (imperfectly) able to discriminate between the same pair of states of  \n the object of interest. We show that\, in principle\, (1) each such pair has a  \n range of possible performance when used optimally (2) in general\, two  \n sensors\, even 'good ones' cannot be combined to be perfect and (3) every  \n sensor has\, in principle\, a magic partner sensor such that the two of them  \n together can\, if the relation between them is at its limit\, be perfect. We  \n speculate briefly on the role of this result in technological and natural  \n systems\, and its implications for the discovery of mechanisms in the natural  \n sciences. Research supported in part by the NSF and by the AFOSR. Paul B.  \n Kantor Biography: Paul Kantor's research centers on the role of information  \n systems for storage and retrieval in a wide range of applications\, with  \n particular emphasis on rigorous evaluation of the effectiveness of such  \n systems. At Rutgers he is a member of the Department of Library and  \n Information Science\, and Research Director of the CCICADA Center. He is also  \n a member of the graduate faculty of the Center for Operations Research  \n (RUTCOR)\,and of the Department of Computer Science\, and is a member of the  \n Center for Discrete Mathematics and Computer Sciences (DIMACS) He is a member  \n of the American Society for Information Science and Technology (ASIST)\, the  \n American Association for the Advancement of Science (AAAS)\, the IEEE\, the  \n American Physical Society\, and the American Statistical Association. His  \n research has been supported by such agencies as the NSF\, DARPA\, ARDA and the  \n US Department of Education. He was educated in Physics and Mathematics at  \n Columbia and Princeton\, has received the ASIST Research award\, and is a  \n Fellow of the AAAS.
END:VEVENT
BEGIN:VEVENT
UID:calendar.10625.field_date.0.263
SUMMARY:Mark Silberstein: Operating system support for high-throughput processors
DTSTAMP:20130618T164412Z
DTSTART:20130204T220000Z
DTEND:20130204T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/operating-system-support-high-throughput-processors
LOCATION:2310 CS
DESCRIPTION:The processor landscape has fractured into latency-optimized CPUs\,  \n throughput-oriented GPUs\, and soon\, custom accelerators. Future applications  \n wwill need to cohesively use a variety of hardware to achieve their  \n performance and power goals. However building efficient systems that use  \n accelerators today is incredibly difficult. In this talk we will argue that  \n the root cause of this complexity lies in the lack of adequate operating  \n system support for accelerators. While operating systems provide optimized  \n resource management and Input/Output (I/O) services to CPU applications\, they  \n make no such services available to accelerator programs. We propose GPUfs -  \n an operating system layer which enables access to files directly from  \n programs running on throughput-oriented accelerators\, such as GPUs. GPUfs  \n extends the constrained GPU-as-coprocessor programming model\, turning GPUs  \n into first-class computing devices with full file I/O support. It provides a  \n POSIX-like API for GPU programs\, exploits parallelism for efficiency\, and  \n optimizes for access locality by extending a CPU buffer cache into physical  \n memories of all GPUs in a single machine. Using real benchmarks we show that  \n GPUfs simplifies the development of efficient applications by eliminating the  \n GPU management complexity\, and broadens the range of applications that can be  \n accelerated by GPUs. For example\, a simple self-contained GPU program which  \n searches for a set of strings in the entire tree of Linux kernel source files  \n completes in about third of the time of an 8-CPU-core run. Joint work with  \n Idit Keidar (Technion)\, Bryan Ford (Yale) and Emmett Witchel (UT Austin) Bio:  \n Mark Silberstein is a postdoctoral fellow in the Operating Systems and  \n Architecture group at the University of Texas at Austin He holds a PhD in  \n Computer Science from the Technion\, Israel. His thesis focused on parallel  \n algorithms and resource management in high-performance large-scale  \n distributed systems. His research in GPU computing includes software-managed  \n caching and memory-intensive applications\, power efficient and hard real-time  \n scheduling in CPU-GPU hybrids\, GPUs and OS privacy\, operating system  \n abstractions and I/O services for GPUs. More information:  \n https://sites.google.com/site/silbersteinmark
END:VEVENT
BEGIN:VEVENT
UID:calendar.11338.field_date.0.264
SUMMARY:Jai Menon: Power Struggles: Revisiting the RISC vs. CISC Debate on  \n Contemporary ARM and x86 Architectures
DTSTAMP:20130618T164412Z
DTSTART:20130212T220000Z
DTEND:20130212T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/power-struggles-revisiting-risc-vs-cisc-debate-contemporary-arm-and-x86-architectures
LOCATION:CS 1221
DESCRIPTION:HPCA Practice talk RISC vs. CISC wars raged in the 1980s when chip area and  \n processor design complexity were the primary constraints and desktops and  \n servers exclusively dominated the computing land- scape. Today\, energy and  \n power are the primary design con- straints and the computing landscape is  \n significantly different: growth in tablets and smartphones running ARM (a  \n RISC ISA) is surpassing that of desktops and laptops running x86 (a CISC  \n ISA). Further\, the traditionally low-power ARM ISA is enter- ing the  \n high-performance server market\, while the traditionally high-performance x86  \n ISA is entering the mobile low-power de- vice market. Thus\, the question of  \n whether ISA plays an intrinsic role in performance or energy efficiency is  \n becoming important\, and we seek to answer this question through a detailed  \n mea- surement based study on real hardware running real applica- tions. We  \n analyze measurements on the ARM Cortex-A8 and Cortex-A9 and Intel Atom and  \n Sandybridge i7 microprocessors over workloads spanning mobile\, desktop\, and  \n server comput- ing. Our methodical investigation demonstrates the role of ISA  \n in modern microprocessors’ performance and energy efficiency. We find that  \n ARM and x86 processors are simply engineering design points optimized for  \n different levels of performance\, and there is nothing fundamentally more  \n energy efficient in one ISA class or the other. The ISA being RISC or CISC  \n seems irrelevant
END:VEVENT
BEGIN:VEVENT
UID:calendar.11518.field_date.0.265
SUMMARY:Nathanael Fillmore: Evaluation of de novo Transcriptome Assemblies from  \n RNA-Seq Data
DTSTAMP:20130618T164412Z
DTSTART:20130212T220000Z
DTEND:20130212T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/evaluation-de-novo-transcriptome-assemblies-rna-seq-data
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) seminar: In order  \n to understand a wide variety of cellular processes\, it is useful to study a  \n cell's transcriptome\, i.e.\, the collection of RNA transcripts present.  \n High-throughput RNA sequencing (RNA-Seq) is a powerful tool toward this end.  \n In order to overcome some of the limitations of current approaches to de novo  \n RNA-Seq assembly\, we have developed a transcriptome assembly scoring function  \n which can be used to choose the best assembly from a collection of candidate  \n assemblies. Our score does not require any knowledge of the organism's genome  \n or of the true set of RNA transcripts. The score is based on a probability  \n model of the process of RNA-Seq read generation and the process of ideal  \n transcriptome assembly. We have also developed several simple reference-based  \n scores\, and we have used these to carry out a large-scale meta-evaluation of  \n our scoring function on real and simulated data.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11519.field_date.0.266
SUMMARY:Catherine Vincent: The Utilization of Machine Learning for the Prediction of  \n Sequence-Informative Fragment Ions Greatly Improves Peptide Identification by  \n Mass Spectrometry
DTSTAMP:20130618T164412Z
DTSTART:20130212T220000Z
DTEND:20130212T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/utilization-machine-learning-prediction-sequence-informative-fragment-ions-greatly-improves-pe
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) seminar: In order  \n to understand a wide variety of cellular processes\, it is useful to study a  \n cell's transcriptome\, i.e.\, the collection of RNA transcripts present.  \n High-throughput RNA sequencing (RNA-Seq) is a powerful tool toward this end.  \n In order to overcome some of the limitations of current approaches to de novo  \n RNA-Seq assembly\, we have developed a transcriptome assembly scoring function  \n which can be used to choose the best assembly from a collection of candidate  \n assemblies. Our score does not require any knowledge of the organism's genome  \n or of the true set of RNA transcripts. The score is based on a probability  \n model of the process of RNA-Seq read generation and the process of ideal  \n transcriptome assembly. We have also developed several simple reference-based  \n scores\, and we have used these to carry out a large-scale meta-evaluation of  \n our scoring function on real and simulated data.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11542.field_date.0.267
SUMMARY:Rathijit Sen: Coordinated Resource Management for Power-Performance  \n Efficiency
DTSTAMP:20130618T164412Z
DTSTART:20130219T190000Z
DTEND:20130219T210000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/coordinated-resource-management-power-performance-efficiency
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: David Wood (advisor)\; Mark Hill\; Somesh Jha\; Mikko Lipasti\;  \n Gurindar Sohi\; Michael Swift
END:VEVENT
BEGIN:VEVENT
UID:calendar.11510.field_date.0.268
SUMMARY:Andrew Nere: HPCA Practice talk : Bridging the Semantic Gap: Emulating  \n Biological Neuronal Behaviors with Simple Digital Neurons
DTSTAMP:20130618T164412Z
DTSTART:20130219T220000Z
DTEND:20130219T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/hpca-practice-talk-bridging-semantic-gap-emulating-biological-neuronal-behaviors-simple-digita
LOCATION:CS 1240
DESCRIPTION:The advent of non von Neumann computational models\, specifically  \n neuromorphic architectures\, has engendered a new class of challenges for  \n computer architects. On the one hand\, each neuron-like computational element  \n must consume minimal power and area to enable scaling up to biological scales  \n of billions of neurons\; this rules out direct support for complex and  \n expensive features like floating point and transcendental functions. On the  \n other hand\, to fully benefit from cortical properties and operations\,  \n neuromorphic architectures must support complex non-linear neuronal  \n behaviors. This semantic gap between the simple and power-efficient  \n processing elements and complex neuronal behaviors has rekindled a RISC vs.  \n CISC-like debate within the neuromorphic hardware design community. In this  \n paper\, we address the aforementioned semantic gap for a recently-described  \n digital neuromorphic architecture that constitutes simple Linear-Leak  \n Integrate-and-Fire (LLIF) spiking neurons as processing primitives. We show  \n that despite the simplicity of LLIF primitives\, a broad class of complex  \n neuronal behaviors can be emulated by composing assemblies of such primitives  \n with low area and power overheads. Furthermore\, we demonstrate that for the  \n LLIF primitives without built-in mechanisms for synaptic plasticity\, two  \n well-known neural learning rules–spike timing dependent plasticity and  \n Hebbian learning–can be emulated via assemblies of LLIF primitives. By  \n bridging the semantic gap for one such system we enable neuromorphic system  \n developers\, in general\, to keep their hardware design simple and  \n power-efficient and at the same time enjoy the benefits of complex neuronal  \n behaviors essential for robust and accurate cortical simulation.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11540.field_date.0.269
SUMMARY:Jeremy C. Weiss: Statistical Timeline Analysis: Clinical Event Prediction  \n from Electronic Health Records
DTSTAMP:20130618T164412Z
DTSTART:20130219T220000Z
DTEND:20130219T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/statistical-timeline-analysis-clinical-event-prediction-electronic-health-records
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) Seminar: Accurate  \n prediction of future onset of disease from Electronic Health Records (EHRs)  \n has important clinical implications. The arrival times of clinical events  \n come at semi-irregular intervals and makes the prediction task challenging.  \n In this talk I discuss two statistical machine learning methods that predict  \n future onset of myocardial infarction\, or heart attack. The first leverages  \n the relational nature of EHRs. The second provides an explicit representation  \n of continuous time\, capturing the irregular arrival times of clinical events.  \n I will discuss the relative advantages of each method and how these ideas may  \n help shape the future customization of health care. function on real and  \n simulated data.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11573.field_date.0.270
SUMMARY:Nicholas A. Davis: The Unique Challenges of Information Security in  \n Healthcare Environments
DTSTAMP:20130618T164412Z
DTSTART:20130226T220000Z
DTEND:20130226T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/unique-challenges-information-security-healthcare-environments
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) Seminar:  \n Information Security is a topic which seems to appear almost daily in the  \n national news. Whether it is Social Security numbers compromised at a  \n prestigious university\, or credit card information\, harvested from a major  \n retailer\, the issues associated with electronic information management have  \n become part of our daily lives. For those people who work in a healthcare  \n environment\, the potential impact of computer security breaches have even  \n greater consequences\, not just in terms of privacy\, but also in the overall  \n quality of healthcare treatment delivered to the patient. The terms  \n confidentiality\, integrity and availability take on a higher level of  \n importance in the healthcare information technology field. This presentation  \n will address both general information security challenges and industry best  \n practices associated with managing\, accessing\, utilizing and protecting  \n electronic healthcare information\, and how to help ensure that patient  \n information remains confidential\, accurate and available to those who need  \n it. During the session\, we will cover technical controls (from a high level)\,  \n as well as process and administrative controls which should be used to  \n protect sensitive data in a healthcare environment. We will also cover such  \n topics as Social Engineering\, which accounts for more than 50% of data  \n security breaches in the United States\, each year. You will learn practical  \n information and skills\, related to how you can best protect PHI and how the  \n organization can work to ensure HIPAA compliance\, from an information  \n technology perspective. Most importantly\, the session will have a Q&A  \n portion\, during which the audience is encouraged to ask specific information  \n technology security questions\, and discuss situations of concern\, related to  \n computer security and electronic healthcare information access and  \n management.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11588.field_date.0.271
SUMMARY:Goetz Graefe: New algorithms for join and grouping operations
DTSTAMP:20130618T164412Z
DTSTART:20130304T181500Z
DTEND:20130304T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/new-algorithms-join-and-grouping-operations
LOCATION:CS 3310
DESCRIPTION:Abstract: Traditional database query processing relies on three types of  \n algorithms for join and for grouping operations. For joins\, index nested  \n loops join exploits an index on its inner input\, merge join exploits sorted  \n inputs\, and hash join exploits differences in the sizes of the join inputs.  \n For grouping\, an index-based algorithm has been used in the past whereas  \n today sort- and hash-based algorithms prevail. Cost-based query optimization  \n chooses the most appropriate algorithm for each query and for each operation.  \n Unfortunately\, mistaken algorithm choices during compile-time query  \n optimization are common yet expensive to investigate and to resolve. Our goal  \n is to end mistaken choices among join algorithms and among grouping  \n algorithms by replacing the three traditional types of algorithms with a  \n single one. Like merge join\, this new join algorithm exploits sorted inputs.  \n Like hash join\, it exploits different input sizes for unsorted inputs. In  \n fact\, for unsorted inputs\, the cost functions for recursive hash join and for  \n hybrid hash join have guided our search for the new join algorithm. In  \n consequence\, the new join algorithm can replace both merge join and hash join  \n in a database management system. The in-memory components of the new join  \n algorithm employ indexes. If the database contains indexes for one (or both)  \n of the inputs\, the new join can exploit persistent indexes instead of  \n temporary in-memory indexes. Using database indexes to find matching input  \n records\, the new join algorithm can also replace index nested loops join. In  \n addition to join operations\, a very similar algorithm supports grouping  \n (“group by” queries in SQL) and duplicate elimination. For unsorted  \n inputs\, candidate output records take on the role of one of the inputs in a  \n join operation. Our goal is to define a single grouping algorithm that can  \n replace grouping by repeated index searches\, by sorting\, and by hashing. In  \n other words\, our goal is to end mistaken algorithm choices not only for joins  \n and other binary matching operations but also for grouping and other unary  \n matching operations in database query processing. Finally\, these new  \n algorithms can be instrumental for efficient and robust data processing in a  \n map-reduce environment\, because “map” and “reduce” operations are  \n similar in essentials to join and grouping operations.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11595.field_date.0.272
SUMMARY:Sriraam Natarajan: Statistical Relational Learning for Predictive  \n Personalized Medicine
DTSTAMP:20130618T164412Z
DTSTART:20130305T220000Z
DTEND:20130305T230000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/statistical-relational-learning-predictive-personalized-medicine
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) Seminar: Recent  \n advances in medicine and electronic book-keeping have greatly increased the  \n amount of medical data available for research and clinical decision making.  \n Electronic Health Records include information about test results\, lab  \n reports\, medical images\, genomics\, treatments\, outcomes\, and family  \n histories. Together with recent advances in data mining and machine learning\,  \n it now seems possible to realize the grand vision of predictive personalized  \n medicine. Statistical Relational Learning (SRL) combines the powerful  \n formalisms of probability theory and first-order logic to handle uncertainty  \n in large\, complex problems. In this talk\, I illustrate the potential of SRL  \n to achieve an important sub-goal of predictive medicine: early detection.  \n Specifically\, I will present SRL approaches for (1) identifying young adults  \n who are at high risk of developing Coronary Heart Disease in middle and later  \n life\, and (2) identifying the set of patients who have or will have  \n Alzheimer's Disease by analyzing their brain MRI images. I will present a  \n general approach for learning SRL models based on Functional-Gradient  \n Boosting. I will adapt this algorithm for the above mentioned challenging  \n tasks to produce state-of-the-art results in three real-world medical  \n studies. I will outline other interesting problems in personalized medicine  \n that we are addressing using SRL and conclude on the optimistic note that  \n predictive personalized medicine is within reach in the near future.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11589.field_date.0.273
SUMMARY:Allison Salmon: WACM Explains... Life in the Commercial Video Game Industry
DTSTAMP:20130618T164412Z
DTSTART:20130306T233000Z
DTEND:20130306T233000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/wacm-explains-life-commercial-video-game-industry
LOCATION:CS 1221
DESCRIPTION:Are you considering a career in video game design? Come meet Allison Salmon\,  \n a senior software engineer and veteran of the commercial game industry.  \n During her years with Raven Software and Activision she shipped seven  \n triple-A titles\, including Quake 4\, Marvel Ultimate Alliance\, and Call of  \n Duty: Black Ops. Allison is now applying her skills to help redefine how  \n video games are used in research and education with her work at the Games\,  \n Learning and Society group at the University of Wisconsin - Madison. This  \n event is hosted by WACM (the Women in Computer Science) and open to all.  \n Oreos will be provided.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11587.field_date.0.274
SUMMARY:Andy Pavlo: Everything I Know About Fast Databases I Learned at the Dog Track
DTSTAMP:20130618T164412Z
DTSTART:20130307T180000Z
DTEND:20130307T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/everything-i-know-about-fast-databases-i-learned-dog-track
LOCATION:CS 2310
DESCRIPTION:Abstract: An emerging class of distributed database management systems  \n (DBMS)\, known as NewSQL\, provide the same scalable performance of NoSQL  \n systems while maintaining the consistency guarantees of a traditional\,  \n single-node DBMS. These NewSQL systems achieve high throughput rates for  \n data-intensive applications by storing their databases in a cluster of main  \n memory partitions. This partitioning enables them to eschew much of the  \n legacy\, disk-oriented architecture that slows down traditional systems\, such  \n as heavy-weight concurrency control algorithms\, thereby allowing for the  \n efficient execution of single-node transactions. But many applications cannot  \n be partitioned such that all of their transactions execute in this manner\;  \n these multi-node transactions require expensive coordination that inhibits  \n performance. Thus\, without intelligent methods to overcome these impediments\,  \n a NewSQL DBMS will scale no better than a traditional DBMS. In this talk\, I  \n will present our research on integrating machine learning techniques to  \n improve the performance of fast database systems that is inspired by my  \n adventures at greyhound racing tracks. In particular\, I will discuss my work  \n on the H-Store parallel\, main memory transaction processing system. I will  \n first describe the Houdini framework that uses Markov models to predict  \n transactions’ behaviors to allow a DBMS to selectively enable runtime  \n optimizations. I will then present Hermes\, a method for the deterministic  \n execution of speculative transactions whenever a DBMS stalls because of  \n distributed transactions. Together\, these projects enable H-Store to support  \n transactional workloads that are beyond what single-node systems can handle.  \n Biography: Andy Pavlo is Ph.D. candidate at Brown University working on  \n database management systems under the circumspect guidance of Stan Zdonik and  \n Michael Stonebraker. His most recent work is focused on the research and  \n development of the H-Store distributed transaction processing system (since  \n commercialized as VoltDB). Before this\, he was a systems programmer for the  \n Condor Project at the University of Wisconsin–Madison with Miron Livny.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11594.field_date.0.275
SUMMARY:Gagan Gupta: Parallel Execution of Ordered Programs on Muliprocessors
DTSTAMP:20130618T164412Z
DTSTART:20130307T190000Z
DTEND:20130307T210000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/parallel-execution-ordered-programs-muliprocessors
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: Gurindar Sohi (advisor)\; Mikko Lipasti\; Kathikeyan Sankaralingam\;  \n Michael Swift\; David Wood
END:VEVENT
BEGIN:VEVENT
UID:calendar.11596.field_date.0.276
SUMMARY:Na Li (Lina): Distributed Energy Management in Power Networks
DTSTAMP:20130618T164412Z
DTSTART:20130308T180000Z
DTEND:20130308T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/distributed-energy-management-power-networks
LOCATION:Engineering Hall\, Room...
DESCRIPTION:The power network is undergoing a fundamental transformation. Future smart  \n grid\, especially on the distribution system\, will be a large-scale network of  \n distributed energy resources (DERs)\, each introducing random and rapid  \n fluctuations in power supply\, demand\, voltage and frequency. These DERs  \n provide tremendous opportunity for sustainability\, efficiency\, and power  \n reliability. However\, there are daunting technical challenges in managing  \n these DERs and optimizing their operation. In this talk\, I will focus on how  \n to develop scalable\, distributed\, and real-time control and optimization to  \n achieve system-wide efficiency\, reliability\, and robustness for the future  \n power grid. In particular I will present how to explore the power network  \n structure to design efficient and distributed market and algorithms for the  \n energy management. I will also talk about how to connect the algorithms with  \n physical dynamics and existing control mechanisms for the real-time control  \n in power networks. Bio: Na Li is currently a Ph.D. candidate in Control and  \n Dynamical Systems at California Institute of Technology. Before that\, she  \n received her B.S. degree at Department of Mathematics\, from ZheJiang  \n University in China in 2007. She was awarded the best student paper award  \n finalist in 2011 IEEE Control and Decision Conference.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11598.field_date.0.277
SUMMARY:James M. Rehg: Behavior Imaging and the Study of Autism
DTSTAMP:20130618T164412Z
DTSTART:20130308T180000Z
DTEND:20130308T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/behavior-imaging-and-study-autism
LOCATION:140 Bardeen
DESCRIPTION:BMI seminar announcement Title: Behavior Imaging and the Study of Autism  \n Speaker: Jim Rehg (Georgia Tech) Location: 140 Bardeen (12:00 noon\, 3/8/2012)  \n Abstract: In this talk I will describe current research efforts in Behavior  \n Imaging\, a new research field which encompasses the measurement\, modeling\,  \n analysis\, and visualization of social and communicative behaviors from  \n multi-modal sensor data. Beginning in infancy\, individuals acquire the social  \n and communicative skills which are vital for a healthy and productive life\,  \n through face-to-face interactions with caregivers and peers. However\,  \n children with developmental delays face great challenges in acquiring these  \n skills\, resulting in substantial lifetime risks. Autism\, for example\, affects  \n 1 in 110 children in the U.S. and can lead to substantial impairments\,  \n resulting in a lifetime cost of care of $3.2M per person. The goal of our  \n research in Behavior Imaging is to develop computational methods that can  \n support the fine-grained and large-scale measurement and analysis of social  \n behaviors\, with the potential to positively impact diagnosis and treatment. I  \n will present an overview of our research efforts in Behavior Imaging\, with a  \n particular emphasis on the use of computer vision techniques. Specifically\, I  \n will describe a new approach to video analysis based on the concept of  \n temporal causality\, which leverages a novel representation of video events as  \n multiple point processes. Our method provides a new bottom-up approach to  \n video segmentation based on the temporal structure of video events. I will  \n present results for retrieving and categorizing social interactions in  \n collections of real-world video footage. I will also highlight our recent  \n efforts in the semi-supervised recognition of objects and activities from  \n egocentric video. This is joint work with KarthirPrabhakar\, Alireza Fathi\,  \n and Sangmin Oh. Speaker Bio: James M. Rehg (pronounced "ray") is a Professor  \n in the School of Interactive Computing at the Georgia Institute of  \n Technology\, where he is the Director of the Center for Behavior Imaging\,  \n co-Director of the Computational Perception Lab\, and Associate Director of  \n Research in the Center for Robotics and Intelligent Machines. He received his  \n Ph.D.from CMU in 1995 and worked at the Cambridge Research Lab of DEC (and  \n then Compaq) from 1995-2001\, where he managed the computer vision research  \n group. He received the National Science Foundation (NSF) CAREER award in  \n 2001\, and the Raytheon Faculty Fellowship from Georgia Tech in 2005.He and  \n his students have received a number of best paper awards\, including best  \n student paper awards at ICML 2005 and BMVC 2010. Dr. Rehg is active in the  \n organizing committees of the major conferences in computer vision\,  \n most-recently serving as the General co-Chair for IEEE CVPR 2009. He has  \n served on the Editorial Board of the International Journal of Computer Vision  \n since 2004. He has authored more than 100 peer-reviewed scientific papers and  \n holds 23 issued US patents. Dr. Rehg is currently leading a multi-institution  \n effort to develop the science and technology of Behavior Imaging\, funded by  \n an NSF Expedition award (see www.cbs.gatech.edu for details).
END:VEVENT
BEGIN:VEVENT
UID:calendar.11605.field_date.0.278
SUMMARY:Joy Arulraj: Production-Run Software Failure Diagnosis via Hardware  \n Performance Counters
DTSTAMP:20130618T164412Z
DTSTART:20130308T210000Z
DTEND:20130308T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/production-run-software-failure-diagnosis-hardware-performance-counters
LOCATION:1221
DESCRIPTION:This is a practice talk for ASPLOS 2013. The talk will be 20 minutes long  \n with time for questions\, feedback\, and discussion after the talk. Sequential  \n and concurrency bugs are widespread in deployed software. They cause severe  \n failures and huge financial loss during production runs. Tools that diagnose  \n production-run failures with low overhead are needed. The state-of-the-art  \n diagnosis techniques use software instrumentation to sample program  \n properties at run time and use off-line statistical analysis to identify  \n properties most correlated with failures. Although promising\, these  \n techniques suffer from high run-time overhead\, which is sometimes over 100%\,  \n for concurrency-bug failure diagnosis and hence are not suitable for  \n production-run usage. We present PBI\, a system that uses existing hardware  \n performance counters to diagnose production-run failures caused by sequential  \n and concurrency bugs with low overhead. PBI is designed based on several key  \n observations. First\, a few widely supported performance counter events can  \n reflect a wide variety of common software bugs and can be monitored by  \n hardware with almost no overhead. Second\, the counter overflow interrupt  \n supported by existing hardware and operating systems provides a natural and  \n effective mechanism to conduct event sampling at user level. Third\, the noise  \n and non-determinism in interrupt delivery complements well with statistical  \n processing. We evaluate PBI using 13 real-world concurrency and sequential  \n bugs from representative open-source server\, client\, and utility programs\,  \n and 10 bugs from a widely used software-testing benchmark. Quantitatively\,  \n PBI can effectively diagnose failures caused by these bugs with a small  \n overhead that is never higher than 10%. Qualitatively\, PBI does not require  \n any change to software and presents a novel use of existing hardware  \n performance counters.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11593.field_date.0.279
SUMMARY:Kunal Talwar: The Geometry of Approximate Differential Privacy
DTSTAMP:20130618T164412Z
DTSTART:20130311T150000Z
DTEND:20130311T160000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/geometry-approximate-differential-privacy
LOCATION:CS 3310
DESCRIPTION:In this talk\, I will discuss trade-offs between accuracy and privacy in the  \n context of linear queries over histograms. This is a rich class of queries  \n that includes contingency tables and range queries\, and has been a focus of a  \n long line of work. For a set of $d$ linear queries over a database $x \in  \n \R^N$\, we seek to find the differentially private mechanism that has the  \n minimum mean squared error. I will first describe a $O(\log^2 d)$  \n approximation algorithm for this problem\, for the case of  \n $(\eps\,\delta)$-differential privacy. The mechanism is simple\, efficient and  \n adds correlated Gaussian noise to the answers. We prove its approximation  \n guarantee relative to the hereditary discrepancy lower bound of Muthukrishnan  \n and Nikolov\, using tools from convex geometry. We will next consider this  \n question in the case when the number of queries exceeds the number of  \n individuals in the database\, i.e. when $d > n = \|x\|_1$. It is known that  \n better mechanisms exist in this setting. I will then describe an  \n $(\eps\,\delta)$-differentially private mechanism which is optimal up to a  \n $\polylog(d\,N)$ factor for any given query set $A$ and any given upper bound  \n $n$ on $\|x\|_1$. This approximation is achieved by coupling the Gaussian  \n noise addition approach with a linear regression step. We give an analogous  \n result for the $\eps$-differential privacy setting. We also improve on the  \n mean squared error upper bound for answering counting queries on a database  \n of size $n$ by Blum\, Ligett\, and Roth\, and match the lower bound implied by  \n the work of Dinur and Nissim up to logarithmic factors. The connection  \n between hereditary discrepancy and the privacy mechanism also enables us to  \n derive the first polylogarithmic approximation to the hereditary discrepancy  \n of a matrix $A$. This talk is based on joint work with Alex Nikolov and Li  \n Zhang
END:VEVENT
BEGIN:VEVENT
UID:calendar.11379.field_date.0.280
SUMMARY:Prof. Kathryn Mckinley: The Yin and Yang of Hardware Heterogeneity: Can  \n Software Survive?
DTSTAMP:20130618T164412Z
DTSTART:20130312T210000Z
DTEND:20130312T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/yin-and-yang-hardware-heterogeneity-can-software-survive
LOCATION:CS 1240
DESCRIPTION:Power and energy constraints are now the driving force in all devices from  \n servers to smartphones. This talk starts with quantitative power\,  \n performance\, and energy measurements on a range of workloads and devices that  \n point to the need for hardware heterogeneity to match software  \n characteristics. However\, programming heterogeneous hardware directly is a  \n nightmare. We show how to abstract\, choose\, and exploit hardware  \n heterogeneity. For interactive server workloads\, we promote jobs from slow to  \n fast cores to deliver substantial improvements in throughput and energy  \n compared to homogeneous designs. These results offer some hope that software  \n may survive and perhaps thrive as heterogeneity hardware evolves in the post  \n Denard era. Speaker Bio Kathryn S. McKinley is a Principal Researcher at  \n Microsoft and an Endowed Professor of Computer Science at The University of  \n Texas at Austin. She and her collaborators have produced widely used tools:  \n the DaCapo Java Benchmarks\, TRIPS Compiler\, Hoard memory manager\, MMTk  \n garbage collector toolkit\, and Immix garbage collector. Her awards include:  \n NSF Career\, ASPLOS 2009 Best Paper\, 2012 IEEE Top Picks\, CACM Research  \n Highlights (2006\, 2012)\, Most Influential OOPSLA Paper from 2002 (awarded  \n 2012)\, the 2011 ACM SIGPLAN Distinguished Service Award\, and the 2012 ACM  \n SIGPLAN Programming Languages Software Award. She has graduated 17 PhD  \n students. She is an IEEE Fellow and ACM Fellow.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11610.field_date.0.281
SUMMARY:Ishmael Amarreh: Characterizing the Etiology and Natural History of  \n Structural and Functional Impairment in Childhood Onset Epilepsy
DTSTAMP:20130618T164412Z
DTSTART:20130312T210000Z
DTEND:20130312T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/characterizing-etiology-and-natural-history-structural-and-functional-impairment-childhood-ons
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) Seminar: Epilepsy  \n and seizures affect almost 3 million Americans of all ages\, at an estimated  \n annual cost of $15.5 billion in direct and indirect costs. In addition\,  \n approximately 200\,000 new cases of seizures and epilepsy occur each year. And  \n out of the ten percent of the American population who will experience a  \n seizure in their lifetime three percent will develop epilepsy by age 75 there  \n are more than 40 types of epileptic syndromes that are defined by repeated  \n epilepstiform discharges or seizures. The prevalence of epilepsy is the  \n highest of all neurological disorders in the world\, and there are an  \n estimated sixty million peoples around the world diagnosed with epilepsy.  \n About 300\,000 American children under the age of 14 have epilepsy. For some  \n epilepsy can be treated with medication and they eventually outgrow it\, but  \n for other\, living with epilepsy is lifelong challenge. Furthermore\, the  \n impact of epilepsy on neurodevelopment and the long-term effects of epilepsy  \n are poorly understood. Defining the effects of epilepsy in the long-term  \n requires illuminating the role and possible interaction of a number of  \n factors that influence this disorder. These factors include altered  \n neurobiological processes that antedate the first recognized seizure  \n epilepsy\, and the changes caused by repeated seizures. This research helped  \n elucidate the state of the brain at onset of epilepsy and the subsequent  \n changes caused by epileptic seizures. The work of this presentation is a part  \n of a longitudinal study of recent-onset idiopathic epilepsy in children and  \n adolescents between the ages of 8 and 18 years at recruitment. Unlike other  \n cross sectional studies of chronic epilepsy with this longitudinal design we  \n were able to separate the antecedent and progressive effects of chronic  \n epilepsy. Three sets of data were collected for each participant with a  \n two-year separation between time one and two and six years between time one  \n and time three. Data collected included a T1\, T2 and DWI MRI scans as well as  \n comprehensive neuropsychological testing. The experiments completed to date  \n involved cross-sectional investigations of time three data of the pediatric  \n epilepsy population. The work has involved developing the post-processing  \n methods for analyzing DTI and structural MRI data.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11608.field_date.0.282
SUMMARY:Muthu Muthukrishnan: Internet Ad Systems and Research Challenges
DTSTAMP:20130618T164412Z
DTSTART:20130314T203000Z
DTEND:20130314T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/internet-ad-systems-and-research-challenges
LOCATION:CS 1240
DESCRIPTION:There is a large\, active\, and effective market for placing ads on the  \n Internet. We will describe currently popular systems for sponsored search as  \n well as the ad exchange for trading display ads on the Internet. We will also  \n present example research problems. Unique problems tend to be on the boundary  \n of Economics\, Auctions\, Game Theory and Computer Science aspects including  \n algorithms\, optimization and data mining.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11597.field_date.0.283
SUMMARY:Javad Lavaei\, Assistant Professor: Low-Rank Solution for Nonlinear  \n Optimization over Graphs
DTSTAMP:20130618T164412Z
DTSTART:20130318T180000Z
DTEND:20130318T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/low-rank-solution-nonlinear-optimization-over-graphs
LOCATION:WID Room 3280 (3rd floor...
DESCRIPTION:In this talk\, we consider an arbitrary real or complex-valued optimization  \n problem whose variable is a positive semi-definite (PSD) matrix. The  \n objective is to investigate under what conditions this optimization has a  \n low-rank (e.g.\, rank 1 or 2) solution. The motivation is that a broad class  \n of optimization problems\, including polynomial optimization\, can be cast as a  \n rank-constrained matrix optimization. To solve the problem\, the structure of  \n the optimization is mapped into a generalized weighted graph\, where each edge  \n is associated with a real/complex weight set. We first show that the problem  \n is NP-hard even in the case when the underlying graph is a tree and each  \n weight set has only two elements. We then introduce the notion of "sign  \n definite real/complex set"\, based on which we prove that the existence of a  \n low-rank solution can be related to the topology of the graph characterizing  \n the optimization as well as the sign definiteness of its weight sets. As a  \n by-product of this result\, several classes of optimizations are shown to be  \n polynomial-time solvable. To demonstrate the application of this result in  \n power systems\, we illustrate that optimization over a power circuit can be  \n mapped into a generalized weighted graph\, where each weight set is naturally  \n sign definite due to the physics of power systems. As another problem related  \n to the topic of this talk\, we finally discuss “generalized network flow  \n (GNF)”. This is a well-studied problem in the special case where the loss  \n of the network is a linear function. Under the reasonable assumption that the  \n loss of each line is an increasing convex function\, we show that GNF is  \n polynomial-time solvable. Biography: Javad Lavaei is an Assistant Professor  \n in the Department of Electrical Engineering at Columbia University. He  \n obtained his Ph.D. degree in Control & Dynamical Systems from California  \n Institute of Technology and held a one-year postdoc position jointly with  \n Electrical Engineering and Precourt Institute for Energy at Stanford  \n University. He is the recipient of the Milton and Francis Clauser Doctoral  \n Prize for the best university-wide Ph.D. thesis. His research interests  \n include power systems\, networking\, distributed computation\, optimization\, and  \n control theory. Javad Lavaei is a senior member of IEEE and has won several  \n awards\, including the Canadian Governor General’s Gold Medal\, Northeastern  \n Association of Graduate Schools Master’s Thesis Award\, New Face of  \n Engineering in 2011\, and Silver Medal in the 1999 International Mathematical  \n Olympiad.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11625.field_date.0.284
SUMMARY:Dave Anderson: WiSDOM Talk: Dave Anderson
DTSTAMP:20130618T164412Z
DTSTART:20130318T180000Z
DTEND:20130318T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/wisdom-talk-dave-anderson
LOCATION:2310
DESCRIPTION:Disk drives today have a set of sophisticated security services well beyond  \n simple data encryption. While these are widely used to support the management  \n of encryption within the drive\, these could conceivably help solve other  \n security problems. This talk will describe the working of drive self  \n encryption and the nature of these security services with the intent of  \n stimulating discussion on potential\, related research projects.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11630.field_date.0.285
SUMMARY:Gary Pack: Semiparametric Geometric Methods for Extracting and Modeling White  \n Matter Volumetric Structures of the Brain
DTSTAMP:20130618T164412Z
DTSTART:20130319T210000Z
DTEND:20130319T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/semiparametric-geometric-methods-extracting-and-modeling-white-matter-volumetric-structures--0
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) Seminar: There is  \n growing evidence that white matter structural systems in the brain have  \n consistent local and global intrinsic geometries. Additionally\, the  \n eigenvectors derived from Diffusion Tensor Imaging data are good indicators  \n of local intrinsic geometry of white matter tissue. This talk describes a  \n novel method for extracting a nonlinear manifold model of the intrinsic  \n geometry of white matter systems of the brain. The Semiparametric Geometric  \n Model takes Diffusion Magnetic Resonance Images as input. Then\, using a  \n combination of semi-supervised learning and semiparametric modeling\,  \n automatically segments white matter structures and outputs a global nonlinear  \n model of the intrinsic geometry of white matter volumes. The manifold model  \n allows the data to be organized and sampled in a way that reflects the  \n physical organization of white matter. For example: Geodesic curves are a  \n robust estimate of long range fiber organization\; manifold surfaces allow  \n better estimation of cross sectional areas and distances\; and manifold  \n volumes are a new\, robust way to estimate white matter connectivity between  \n regions of the brain. Finally\, new methods of statistical and geometric  \n analysis on the white matter structures are presented.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11525.field_date.0.286
SUMMARY:Patrick Traynor: Chasing Telephony Security: Where The Wild Things... Are?
DTSTAMP:20130618T164412Z
DTSTART:20130321T210000Z
DTEND:20130321T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/chasing-telephony-security-where-wild-things-are
LOCATION:1221
DESCRIPTION:Abstract: Mobile phones are now used by over five billion people\, making them  \n the dominant computing platform around the world. This shift has not gone  \n unnoticed - malicious behavior targeting mobile devices has recently been one  \n of the most popular security topics in both academic circles and the popular  \n press. However\, the extent to which such threats are actually present in user  \n devices is not well understood. This talk begins with the first study trying  \n to establish ground truth of mobile malware in a major US cellular provider  \n and offers insight into the reality of such threats. I then focus on the  \n recent surge in Caller-ID spoofing and associated attacks\, and discuss a  \n mechanism to detect such attacks. Bio: Patrick Traynor is an Assistant  \n Professor in the College of Computing at Georgia Tech. His research focuses  \n on the security of mobile systems\, with a concentration on telecommunications  \n infrastructure and mobile devices. His research has uncovered critical  \n vulnerabilities in cellular networks\, made the first characterization of  \n mobile malware in provider networks and offers a robust approach to detecting  \n and combatting Caller-ID scams. He is also interested in Internet security  \n and the systems challenges of applied cryptography. Professor Traynor earned  \n his Ph.D and M.S. in Computer Science and Engineering from the Pennsylvania  \n State University in 2008 and 2004\, respectively\, and his B.S. in Computer  \n Science from the University of Richmond in 2002. He is currently a member of  \n the Georgia Tech Information Security Center (GTISC) and a co-director of the  \n Converging Infrastructure Security Laboratory (CISEC). He is also a  \n co-founder of Pindrop Security.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11607.field_date.0.287
SUMMARY:Mahesh V. Tripunitara: An attack- and a defence-mechanism in the context of  \n hardware security
DTSTAMP:20130618T164412Z
DTSTART:20130326T203000Z
DTEND:20130326T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/attack-and-defence-mechanism-context-hardware-security
LOCATION:CS 2310
DESCRIPTION:With the increased outsourcing of the fabrication of digital Integrated  \n Circuits (ICs)\, security is seen as a concern. The threat agent is someone at  \n the foundry\, perhaps in collusion with a designer or a user\, that maliciously  \n modifies ICs during fabrication\, for example\, by inserting a backdoor. In  \n this talk\, I will discuss two pieces of on-going work in this context. The  \n first piece is the realization and validation of a non-deterministic hardware  \n timer that can be used to trigger a backdoor. Prior work has considered  \n deterministic timers\, i.e.\, those that go off with probability 1\, and has  \n left open issues regarding the effectiveness\, from the standpoint an  \n attacker\, of non-deterministic timers\, i.e.\, those that have a random  \n component. Our work addresses these open issues and shows that such timers  \n can be realized with powerful properties to an attacker\, in a manner that the  \n bar on potential defence mechanisms is raised considerably. The second piece  \n of work I will discuss is a defence-mechanism that leverages 3D IC technology  \n that splits a circuit into multiple tiers\, each of which may be fabricated  \n separately\, and then stacked vertically and connected using Through-Silicon  \n Vias (TSVs). Prior work has proposed that such technology can be used to  \n secure digital ICs\, but provides no technical insight or details on how this  \n would work. I will discuss our work that proposes a concrete way of  \n leveraging such technology for security. This includes a characterization of  \n security\, and the computational complexity of the underlying problem. I will  \n discuss also an approach we have implemented and present empirical results on  \n benchmark circuits\, and a case-study of a circuit for DES. (This is joint  \n work with Frank Imeson and Siddharth Garg of the University of Waterloo.)  \n Bio: Mahesh Tripunitara is an assistant professor in the ECE department at  \n the University of Waterloo in Canada\, where he had been since 2009. He works  \n mostly in information security\, on problems in access control\, conditional  \n payments\, cryptographic key transport and more recently\, computer hardware.  \n He has a PhD in computer science from Purdue University\, and about 9 years of  \n industry-experience.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11634.field_date.0.288
SUMMARY:Andrew Moore: NetFPGA: The Flexible Open-Source Networking Platform
DTSTAMP:20130618T164412Z
DTSTART:20130402T160000Z
DTEND:20130402T171500Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/netfpga-flexible-open-source-networking-platform
LOCATION:4310
DESCRIPTION:Abstract: The NetFPGA is an open platform enabling researchers and  \n instructors to build high-speed\, hardware-accelerated networking systems. The  \n NetFPGA is the de-facto experimental platform for line-rate implementations  \n of network research and it continues with a new generation platform capable  \n of 4x10Gbps. The target audience is not restricted to hardware researchers:  \n the NetFPGA provides the ideal platform for research across a wide range of  \n networking topics from architecture to algorithms and from energy-efficient  \n design to routing and forwarding. The most prominent NetFPGA success is  \n OpenFlow\, which in turn has reignited the Software Defined Networking  \n movement. NetFPGA enabled OpenFlow by providing a widely available  \n open-source development platform capable of line-rate and was\, until its  \n commercial uptake\, the reference platform for OpenFlow. NetFPGA enables  \n high-impact network research. This seminar will combine presentation and  \n demonstration\; no knowledge of hardware programming languages (eg  \n Verilog/VHDL) is required. A NetFPGA 10G card will be awarded as a door-prize  \n amongst the seminar attendees. Speaker bio: ANDREW W. MOORE is a Senior  \n Lecturer at the University of Cambridge Computer Laboratory in England\, where  \n he is part of the Systems Research Group working on issues of network and  \n computer architecture. His research interests include enabling open-source  \n network research and education using the NetFPGA platform\, other research  \n pursuits include low-power energy-aware networking\, and novel network and  \n systems data-center architectures. He holds B.Comp. and M.Comp. degrees from  \n Monash University and a Ph.D. from the University of Cambridge. He is a  \n chartered engineer with the IET and a member of the IEEE\, ACM and USENIX.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11506.field_date.0.289
SUMMARY:Leng Leng Tan\, Senior Vice President at Oracle: WACM Speaker Series:  \n Venusians in Mars? Thriving Tips for Women in Computing
DTSTAMP:20130618T164412Z
DTSTART:20130404T210000Z
DTEND:20130404T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/wacm-speaker-series-venusians-mars-thriving-tips-women-computing
LOCATION:CS 2310
DESCRIPTION:** This event is open to everyone ** As women in computing\, we sometimes feel  \n like we are Venusians (residents of the planet Venus) dwelling in Mars\,  \n struggling to survive in this male-dominated field. Do we have to change our  \n identity in order to thrive in the computing field? Speaking from her  \n experience\, Leng will discuss the common issues and pitfalls we encounter and  \n she will explore how we can leverage our strengths to excel and lead!  \n Speaker's Bio: Leng Leng Tan is the Senior Vice President of Engineering for  \n Oracle Enterprise Manager product at Oracle Corporation. She was the leader  \n behind the widely adopted self-managing database capabilities in Oracle  \n Database Diagnostic and Tuning Packs. Her team also delivered the innovative  \n workload capture and replay technologies (Real Application Testing) that  \n enable realistic testing of database and middle-tier. The most recent release  \n of Oracle Enterprise Manager 12c Cloud Control provides an integrated system  \n management tool that manages Oracle software and hardware including  \n virtualization and Cloud. Leng was born in Singapore and came to University  \n of Wisconsin\, Madison for her B.S. in computer science. She also holds an  \n M.S. in computer science from Stanford University. Leng originally joined  \n Oracle in 1989 as the sole female database kernel developer.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11646.field_date.0.290
SUMMARY:WACM Explains... Entrepreneurship
DTSTAMP:20130618T164412Z
DTSTART:20130410T230000Z
DTEND:20130411T000000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/wacm-explains-entrepreneurship
LOCATION:CS 1221
DESCRIPTION:Come hear local entrepreneurs talk about their experiences going off the  \n traditional paths and blazing through new territory to make their dreams come  \n true. They'll discuss the challenges and rewards of starting a business\, talk  \n about their journeys\, and answer questions. Moderator: Heather Wentler  \n (Fractal / The Doyenne Group) Panelists: Forrest Woolworth (Capital  \n Entrepreneurs / PerBlue) Jami Morton (SnowShoe) Betsy Rowbottom (The Good  \n Jobs) Tim Kessler (Badger Innovations / Collegiate Entrepreneurs Org.) We  \n will also include time for networking with the panelists after the talk. This  \n event is hosted by WACM (the Women in Computer Science) and open to all.  \n Oreos will be provided.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11654.field_date.0.291
SUMMARY:P. Brighten Godfrey: Veriflow: Verifying Network-Wide Invariants in Real Time
DTSTAMP:20130618T164412Z
DTSTART:20130412T200000Z
DTEND:20130412T210000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/veriflow-verifying-network-wide-invariants-real-time
LOCATION:CS 3310
DESCRIPTION:Abstract: The increasing complexity of modern computer networks has far  \n outpaced the development of tools to manage their operation. We are  \n developing systems which simplify network security and management by formally  \n reasoning about network-wide forwarding behavior. Our first data plane  \n verification system\, Anteater [SIGCOMM'11]\, revealed multiple real-world bugs  \n in a large university network\, including forwarding loops and stale ACL  \n rules. VeriFlow [HotSDN'12 and NSDI'13] checks network-wide invariants in  \n real time as each forwarding rule is inserted\, optionally blocking  \n vulnerabilities from being introduced into the network. Our current  \n OpenFlow-based implementation can perform rigorous checking within hundreds  \n of microseconds per rule insertion. This talk presents work with Ahmed  \n Khurshid\, Haohui Mai\, Kelvin Zou\, Wenxuan Zhou\, Rachit Agarwal\, Matthew  \n Caesar\, and Sam King. Speaker bio: P. Brighten Godfrey is an assistant  \n professor in the Department of Computer Science at the University of Illinois  \n at Urbana-Champaign. He completed his Ph.D. at UC Berkeley in May 2009\,  \n advised by Ion Stoica\, and his B.S. at Carnegie Mellon University in 2002.  \n His research interests lie in the design and analysis of networked systems.  \n He is a winner of the 2012 National Science Foundation CAREER Award\, the 2012  \n IEEE Communications Society & Information Theory Society Joint Paper Award\,  \n the 2010 IEEE Communications Society Data Storage Technical Committee Best  \n Paper Award\, and a best paper award at HotSDN 2012.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11660.field_date.0.292
SUMMARY:Dolores Severtson and Jeff Myers: Modeling Estimated Cancer Risk from Air  \n Pollution & Designing Maps to Communicate the Uncertainty of the Modeled  \n Estimates
DTSTAMP:20130618T164412Z
DTSTART:20130416T210000Z
DTEND:20130416T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/modeling-estimated-cancer-risk-air-pollution-designing-maps-communicate-uncertainty-modeled-es
LOCATION:Biotechnology Center...
END:VEVENT
BEGIN:VEVENT
UID:calendar.11664.field_date.0.293
SUMMARY:Michael Mahoney: Revisiting the Nystrom Method for Improved Large-Scale  \n Machine Learning
DTSTAMP:20130618T164412Z
DTSTART:20130418T210000Z
DTEND:20130418T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/revisiting-nystrom-method-improved-large-scale-machine-learning
LOCATION:CS1240
DESCRIPTION:The Nystrom Method is a sampling-based or CUR-like low-rank approximation to  \n symmetric positive semi-definite (SPSD) matrices such as Laplacians and  \n kernels that arise in machine learning. The method has received a great deal  \n of attention in recent years\, much of which has focused on the use of  \n randomized algorithms to compute efficiently the low-rank approximation.  \n Motivated by conflicting and contradictory claims in the literature\, we have  \n performed a detailed empirical evaluation of the performance quality and  \n running time of sampling-based and projection-basedlow-rank approximation  \n methods on a diverse suite of SPSD matrices. Our main conclusions are  \n threefold: first\, our results highlight complementary aspects of projection  \n methods versus sampling based on the statistical leverage scores\; second\, our  \n results elucidate the effect of popular "design decisions" on the structural  \n properties of the matrices and thus on the applicability of different  \n approximation algorithms\; and third\, our results show that sampling-based  \n algorithms that use provably-accurateapproximate leverage scores can have  \n running times comparable to random projection algorithms. In addition\, our  \n empirical results demonstrate that prior theory was so weak as to not even be  \n a qualitative guide to practice. Thus\, we complement our empirical results  \n with a suite of worst-case theoretical bounds for both random sampling and  \n random projections methods. These bounds are qualitatively superior to  \n existing bounds\, e.g.\, improved additive-error bounds for the spectral and  \n Frobenius norm errors and relative-error bounds for the trace norm error\, and  \n they can be used to understand the complementary aspects of projection versus  \n sampling-based methods. We will also describe how recent numerical  \n implementations of random sampling and random projection algorithms can be  \n used in this setting in much larger-scale applications.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11662.field_date.0.294
SUMMARY:Dr. Nina Taft: Privacy Preserving Ridge Regression on Hundreds of Millions of  \n Records
DTSTAMP:20130618T164412Z
DTSTART:20130423T203000Z
DTEND:20130423T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/privacy-preserving-ridge-regression-hundreds-millions-records
LOCATION:CS2310
DESCRIPTION:Abstract: Numerous data mining tasks\, used inside cloud services such as  \n recommendation systems\, often employ linear or ridge regression as an  \n underlying computational element. In this talk we present a system in which  \n ridge regression (that includes linear regression) is carried out in a  \n privacy preserving way because the user data stays encrypted all the time.  \n Ridge regression is an algorithm that takes as input a large number of data  \n points and finds the best-fit linear curve through them. Our system outputs  \n the best-fit curve in the clear\, but exposes no other information. We propose  \n a hybrid approach that combines homomorphic encryption with Yao garbled  \n circuits. Our system scales nicely because we remove the dependency on the  \n number of users from any computations involving non-linear operations. We  \n implement the complete system and experiment with it on real data-sets\, and  \n show that our hybrid approach performs significantly better than either  \n method alone. We demonstrate that we can run regression on millions of users'  \n data within a few minutes\, outperforming the state of the art by 2 orders of  \n magnitude. Showing that core data mining building blocks can indeed be  \n executed quickly on encrypted data\, even when there are millions of users in  \n the system\, is an important step in bring privacy to data mining driven cloud  \n services. Bio: Nina Taft received her PhD from UC Berkeley\, and has spent her  \n career working in industrial research labs in the San Francisco Bay Area. She  \n spent 5 years working at Sprint Labs in the IP research group that helped  \n launch the field of Internet Measurement. Nina worked on ISP traffic  \n engineering problems\, including capacity planning and routing\, but is  \n primarily known for her body of work on traffic matrices. After Sprint\, Nina  \n worked for Intel Labs in Berkeley and conducted research on anomaly  \n detection\, energy management\, and end-host tracing tools for automated  \n performance diagnosis. Currently she is a distinguished scientist at  \n Technicolor Research Palo Alto where she focuses on privacy and  \n recommendation systems. Nina is an active member of the networking community  \n where she has served on numerous program committees\, steering committees and  \n in various conference chair positions.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11667.field_date.0.295
SUMMARY:Jeremy Goecks\, PhD: Interactive Computing for High-Performance Medical  \n Genomics
DTSTAMP:20130618T164412Z
DTSTART:20130423T210000Z
DTEND:20130423T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/interactive-computing-high-performance-medical-genomics
LOCATION:Biotechnology Center...
DESCRIPTION:Computation and Informatics in Biology and Medicine (CIBM) Seminar: Low-cost\,  \n high-throughput DNA sequencing has become widespread and is revolutionizing  \n biomedical research and clinical care alike. In the era of pervasive  \n genomics\, the greatest challenge is making sense of large sequencing  \n datasets. The Galaxy platform (http://galaxyproject.org\,  \n http://usegalaxy.org) is a popular Web-based workbench that enables  \n accessible\, reproducible\, and collaborative analysis of genomic data. Galaxy  \n makes it easy for anyone\, regardless of their programming experience\, to  \n analyze large genomic datasets. Galaxys visual analytics framework advances  \n the state-of-the-art in interactive high-performance genomics\, enabling  \n investigators to simultaneously use visualization and scientific tools to  \n analyze genomic data. In collaboration with the Emory Winship Cancer  \n Institute\, we are using Galaxy to analyze cancer genomes and uncover  \n mutation-gene expression associations in cancer. Knowledge derived from these  \n analyses can be used to improve outcomes for cancer patients\, such as by  \n determining which patients are most likely to suffer a relapse. As part of  \n this research\, we have identified opportunities for extending Galaxy to  \n enable large\, integrated clinical informatics applications.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11680.field_date.0.296
SUMMARY:Laura Legault: Analysis of Alternative Splicing Using Probabilistic Splice  \n Graphs and RNA Sequencing
DTSTAMP:20130618T164412Z
DTSTART:20130424T130000Z
DTEND:20130424T150000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/analysis-alternative-splicing-using-probabilistic-splice-graphs-and-rna-sequencing
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: Colin Dewey (advisor) Sushmita Roy David Wasserman Aaron Hoskins
END:VEVENT
BEGIN:VEVENT
UID:calendar.11681.field_date.0.297
SUMMARY:Aditya Thakur: From Machine-Code Verification to Satisfiability Modulo  \n Abstraction
DTSTAMP:20130618T164412Z
DTSTART:20130425T150000Z
DTEND:20130425T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/machine-code-verification-satisfiability-modulo-abstraction
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: Tom Reps (advisor) Somesh Jhaa Ken McMIllan Mooly Sagiv
END:VEVENT
BEGIN:VEVENT
UID:calendar.11658.field_date.0.298
SUMMARY:Irene Rae & Dan Szafir: CHI Practice Talks
DTSTAMP:20130618T164412Z
DTSTART:20130425T180000Z
DTEND:20130425T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/chi-practice-talks
LOCATION:1240 CS
DESCRIPTION:Irene Rae and Dan Szafir\, graduate students in the HCI Group\, will be giving  \n practice talks for their upcoming presentations at the CHI Conference. The  \n HCI Lab invites graduate students and faculty to come enjoy cookies\, hear  \n about Irene and Dan's research\, and give them feedback. Talk 1: In-body  \n Experiences: Embodiment\, Control\, and Trust in Robot-Mediated Communication  \n Irene Rae\, Leila Takayama\, & Bilge Mutlu Communication technologies are  \n becoming increasingly diverse in form and functionality\, making it important  \n to identify which aspects of these technologies actually improve  \n geographically distributed communication. Our study examines two potentially  \n important aspects of communication technologies which appear in  \n robot-mediated communication--physical embodiment and control of this  \n embodiment. We studied the impact of physical embodiment and control upon  \n interpersonal trust in a controlled laboratory experiment using three  \n different videoconferencing settings: (1) a handheld tablet controlled by a  \n local user\, (2) an embodied system controlled by a local user\, and (3) an  \n embodied system controlled by a remote user (n = 29 dyads). We found that  \n physical embodiment and control by the local user increased the amount of  \n trust built between partners. These results suggest that both physical  \n embodiment and control of the system influence interpersonal trust in  \n mediated communication and have implications for future system designs. Talk  \n 2: ARTFuL: Adaptive Review Technology for Flipped Learning Daniel Szafir\,  \n Bilge Mutlu Internet technology is revolutionizing education. Teachers are  \n developing massive open online courses (MOOCs) and using innovative practices  \n such as flipped learning in which students watch lectures at home and engage  \n in hands-on\, problem solving activities in class. This work seeks to explore  \n the design space afforded by these novel educational paradigms and to develop  \n technology for improving student learning. Our design\, based on the technique  \n of adaptive content review\, monitors student attention during educational  \n presentations and determines which lecture topic students might benefit the  \n most from reviewing. An evaluation of our technology within the context of an  \n online art history lesson demonstrated that adaptively reviewing lesson  \n content improved student recall abilities 29% over a baseline system and was  \n able to match recall gains achieved by a full lesson review in less time. Our  \n findings offer guidelines for a novel design space in dynamic educational  \n technology that might support both teachers and online tutoring systems.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11635.field_date.0.299
SUMMARY:William H Sandholm\, Professor: Population Games and Evolutionary Dynamics
DTSTAMP:20130618T164412Z
DTSTART:20130426T170000Z
DTEND:20130426T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/population-games-and-evolutionary-dynamics
LOCATION:WID\, room 3280 (3rd...
DESCRIPTION:Population games provide a general model of strategic interactions among  \n large numbers of agents\; highway congestion\, multilateral externalities\, and  \n natural selection are among their many applications. To model the dynamics of  \n behavior in population games\, we introduce decision protocols\, which provide  \n explicit stochastic descriptions of how individual agents make decisions.  \n When the number of agents is large enough\, the evolution of aggregate  \n behavior can be described by solutions to ordinary differential equations. We  \n discuss classes of population games in which these evolutionary dynamics lead  \n to equilibrium play\, we consider simple examples in which cycling and chaos  \n can arise\, and we explain how natural decision protocols can generate potent  \n equilibrium selection results. Finally\, we discuss computational experiments  \n whose aim is to estimate the frequency of cycling and chaos in randomly  \n chosen games.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11682.field_date.0.300
SUMMARY:Sriram Subramanian: Beyond the Block-Based Interface for Flash-Based Storage
DTSTAMP:20130618T164412Z
DTSTART:20130426T193000Z
DTEND:20130426T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/beyond-block-based-interface-flash-based-storage
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: Andrea Arpaci-Dusseau (co-advisor) Remzi Arpaci-Dusseau  \n ((co-advisor) Shan Lu Mike Swift Jon Eckhardt
END:VEVENT
BEGIN:VEVENT
UID:calendar.11677.field_date.0.301
SUMMARY:Marc Snir: Resilience at Exascale
DTSTAMP:20130618T164412Z
DTSTART:20130501T160000Z
DTEND:20130501T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/resilience-exascale
LOCATION:H.F. DeLuca Forum\,...
DESCRIPTION:It is often feared that the growing frequency of hardware errors will be a  \n major obstacle to the deployment of exascale systems. The Department of  \n Energy held several workshops to study this issue\, including a week-long  \n workshop organized by the Institute for Computing Sciences. This workshop  \n brought together leading researchers in circuits\, architecture\, operating  \n systems and applications. The workshop produced a report that indicates  \n possible scenarios for handling resilience at exascale and required research  \n to achieve progress in this area. Snir will discuss this report\, indicating  \n the questions it raises and research directions it identifies. Marc Snir is  \n Director of the Mathematics and Computer Science Division at the Argonne  \n National Laboratory and Michael Faiman and Saburo Muroga Professor in the  \n Department of Computer Science at the University of Illinois at  \n Urbana-Champaign. He currently pursues research in parallel computing.He was  \n head of the Computer Science Department from 2001 to 2007. Until 2001 he was  \n a senior manager at the IBM T. J. Watson Research Center where he led the  \n Scalable Parallel Systems research group that was responsible for major  \n contributions to the IBM SP scalable parallel system and to the IBM Blue Gene  \n system.Marc Snir received a Ph.D. in Mathematics from the Hebrew University  \n of Jerusalem in 1979\, worked at NYU on the NYU Ultracomputer project in  \n 1980-1982\, and was at the Hebrew University of Jerusalem in 1982-1986\, before  \n joining IBM. Marc Snir was a major contributor to the design of the Message  \n Passing Interface. He has published numerous papers and given many  \n presentations on computational complexity\, parallel algorithms\, parallel  \n architectures\, interconnection networks\, parallel languages and libraries and  \n parallel programming environments.Marc is Argonne Distinguished Fellow\, AAAS  \n Fellow\, ACM Fellow and IEEE Fellow. He has Erdos number 2 and is a  \n mathematical descendant of Jacques Salomon Hadamard. Hosted by the Center for  \n High Throughput Computing
END:VEVENT
BEGIN:VEVENT
UID:calendar.11668.field_date.0.302
SUMMARY:Lingjia Tang: Mitigating Resource Contention in Warehouse-scale Computers
DTSTAMP:20130618T164412Z
DTSTART:20130501T210000Z
DTEND:20130501T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/mitigating-resource-contention-warehouse-scale-computers
LOCATION:1240 CS
DESCRIPTION:Modern datacenters that host large-scale Internet services are extremely  \n expensive to construct and operate. Improving software performance and server  \n utilization is key to improving the efficiency and reducing the enormous cost  \n in datacenters. In this talk\, I present novel compilation techniques and  \n runtime systems to significantly improve performance\, quality of service  \n (QoS) and machine utilization in datacenters by effectively mitigating memory  \n resource contention on modern multicore servers. Specifically\, this talk  \n presents: 1) comprehensive characterization of the impact of memory resource  \n sharing on industry-strength large-scale datacenter workloads and the design  \n of runtime systems to intelligently map application threads to cores to  \n promote positive resource sharing and mitigate resource contention to improve  \n application performance\; 2) the design of novel compilation techniques and  \n run-time systems that statically and dynamically manipulate applications’  \n contentious nature to enable the co-location of applications with varying QoS  \n requirements\, and as a result\, greatly improve server utilization in data  \n centers.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11685.field_date.0.303
SUMMARY:Jason Stowe: HTCondor and Utility HPC in Life Sciences: Use Cases and Lessons  \n Learned
DTSTAMP:20130618T164412Z
DTSTART:20130501T213000Z
DTEND:20130501T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/htcondor-and-utility-hpc-life-sciences-use-cases-and-lessons-learned
LOCATION:H.F. DeLuca Forum\,...
DESCRIPTION:What happens when scientists and researchers are no longer limited by  \n fixed-size compute and data capacity? Better\, faster science. From small  \n research labs to start-ups to Fortune 500s\, the combination of HTCondor and  \n Utility HPC is making impossible science possible in Life Sciences. This  \n session explores how organizations are running large-scale high performance  \n computing (HPC) workloads\, using CycleCloud and HTCondor with lessons learned  \n from several real-world examples in Genomics\, Molecular Modeling\, Simulation\,  \n Proteomics and many more. Jason Stowe\, CEO at Cycle ComputingJason Stowe is a  \n seasoned entrepreneur\, and the founder and CEO of Cycle Computing\, the leader  \n in Utility HPC and Utility Supercomputing Software. Cycle delivers proven\,  \n secure and flexible high performance computing (HPC) and data solutions since  \n 2005. Cycle Computing products help clients maximize internal infrastructure  \n and increase power as research demands\, like the 10000-core cluster for  \n Genentech\, the 30000+ core cluster for a Top 5 Pharma and 50\,000+ core  \n cluster for Schrodinger that were covered in the NY Times\, Wall Street  \n Journal\, Wired\, BusinessWeek\, Bio-IT World and Forbes. Starting with three  \n initial Fortune 100 clients\, Cycle has grown to deploy proven implementations  \n at Fortune 500s\, SMBs and government and academic institutions including JP  \n Morgan Chase\, The Hartford Insurance Group\, Johnson & Johnson\, Purdue  \n University\, Pfizer and Lockheed Martin. Jason attended Carnegie Mellon and  \n Cornell Universities\, and volunteered/guest lectured for the Entrepreneurship  \n program at Cornell's Johnson Business School
END:VEVENT
BEGIN:VEVENT
UID:calendar.11679.field_date.0.304
SUMMARY:Deborah Estrin: Sensemaking for Mobile Health
DTSTAMP:20130618T164412Z
DTSTART:20130502T210000Z
DTEND:20130502T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/sensemaking-mobile-health
LOCATION:Computer Sciences 1240
DESCRIPTION:"Sensemaking for Mobile Health\," explores the possibilities of using data  \n from mobile electronic devices to monitor and manage health care. Speaker  \n Bio: Deborah Estrin is currently on leave from her position as a Professor of  \n Computer Science with a joint appointment in Electrical Engineering at UCLA\,  \n where she held the Jon Postel Chair in Computer Networks\, and was Founding  \n Director of the NSF-funded Center for Embedded Networked Sensing (CENS\,  \n 2001-2012). She has accepted a faculty position with the Computer Science  \n Department at the new Cornell Tech campus in New York City\,  \n http://tech.cornell.edu. Estrin received her Ph.D. (1985) in Computer Science  \n from the Massachusetts Institute of Technology\, and her B.S. (1980) from U.C.  \n Berkeley. Estrin’s early research (conducted while on the Computer Science  \n Department faculty at USC and the USC Information Sciences Institute) focused  \n on the design of network and routing protocols for very large\, global\,  \n networks\, including: multicast routing protocols\, self-configuring protocol  \n mechanisms for scalability and robustness\, and tools and methods for  \n designing and studying large scale networks. In the late 90’s Professor  \n Estrin began her work in embedded networked sensing systems\, with emphasis on  \n environmental monitoring applications. Most recently her work focuses on  \n participatory sensing systems\, leveraging the location\, activity\, image\, and  \n user-contributed data streams increasingly available from mobile phones.  \n Ongoing projects include Participatory Sensing for civic engagement and STEM  \n education (http://mobilizingcs.org)\, and self-monitoring applications in  \n support of health and wellness (http://openmhealth.org).
END:VEVENT
BEGIN:VEVENT
UID:calendar.11728.field_date.0.305
SUMMARY:Ben Liblit: CS 638 [Software Engineering] project demo day
DTSTAMP:20130618T164412Z
DTSTART:20130507T193000Z
DTEND:20130507T204500Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/cs-638-software-engineering-project-demo-day
LOCATION:Unit III first-floor...
DESCRIPTION:CS 638 [Software Engineering] students work together in large semester  \n project teams. Join us any time from 2:30pm until 3:45pm to see live demos  \n of: Intramural Baseball Manager\, proposed by David Kuehn\, developed by Brian  \n Anderson\, Matt Beaty\, David Kuehn\, Colin Laska\, and Kristie Stalberger  \n NexTrack\, proposed by Ryan Riebling\, developed by Aaron Bregger\, William  \n Justmann\, Michael Landau\, Ryan Riebling\, Mikhail Skobov\, and James Stefanich  \n Space Battle Game\, proposed by Colin McKay\, developed by Joseph Francke\,  \n Andrew Hermus\, Pierce Johnson\, Colin McKay\, and Sam Olver Tablet-Top RPG\,  \n proposed by Aaron Bartholomew\, developed by Aaron Bartholomew\, Jacob Laska\,  \n James Merrill\, and Ahmad Faiz Abdull Rashid Tuter: The Ultimate Tutor Finder\,  \n proposed by Sher Minn Chong\, developed by Sher Minn Chong\, Trever Johnson\,  \n Faiz Lurman\, Josh Serbus\, and Adam Thorson UW–Madison Campus Tour Guide\,  \n developed by proposed by Anousone Bounket Anousone Bounket\, Peter Erickson\,  \n Emily Gerner\, Vinodh Muthiah\, and Ryan Shenk These students have all worked  \n very hard\, and accomplished much in little time. Please drop by at your  \n convenience to see and celebrate their accomplishments!
END:VEVENT
BEGIN:VEVENT
UID:calendar.11735.field_date.0.306
SUMMARY:Mohit Saxena: New Interfaces for Solid-State Memory Management
DTSTAMP:20130618T164412Z
DTSTART:20130509T184500Z
DTEND:20130509T204500Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/new-interfaces-solid-state-memory-management
LOCATION:CS 3310
DESCRIPTION:Abstract: The availability of high-speed flash solid-state devices (SSD) has  \n introduced a new tier into the memory hierarchy. SSDs have dramatically  \n different properties than disks\, yet are exposed in many systems as a generic  \n block device. In this talk\, I will present two systems that update the  \n interface to these devices to better match their capabilities as a new memory  \n tier. First\, I will talk about a system called FlashTier to use flash SSDs as  \n a cache in front of slower disks. In this work\, we investigate the numerous  \n differences between the interface offered by an SSD\, a persistent block  \n store\, and the service it provides\, caching data. I will present how we  \n redress these differences through new block-addressing and space-management  \n techniques. Next\, I will describe our work on extending main memory by  \n virtualizing it with inexpensive flash storage. We find that there are  \n several paging mechanisms in the core virtual memory subsystem of the Linux  \n kernel\, which have been optimized for the characteristics of disks. I will  \n describe a new flash-virtual memory system called FlashVM that de-diskifies  \n these mechanisms for improved performance and reliability with flash SSDs.  \n Ph.D Defense Committee Members: Michael M. Swift (chair)\, UW-Madison Andrea  \n C. Arpaci-Dusseau\, UW-Madison Remzi H. Arpaci-Dusseau\, UW-Madison Mark D.  \n Hill\, UW-Madison Arif Merchant\, Google
END:VEVENT
BEGIN:VEVENT
UID:calendar.11669.field_date.0.307
SUMMARY:James F. O'Brien: Perception\, Measurement\, and Simulation
DTSTAMP:20130618T164412Z
DTSTART:20130509T210000Z
DTEND:20130509T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/perception-measurement-and-simulation
LOCATION:Computer Sciences 1240
DESCRIPTION:Abstract: This talk covers several graphics projects including cloth  \n simulation\, perceptually based tone mapping\, and digital image/video  \n forensics. These projects will be discussed in the context of exploring the  \n common underlying theme of simulation based on human perception and  \n measurement. I will show how measured data and models of the human visual  \n system can be used for more realistic image reproduction\, how simulation can  \n be used to take measurements of the real-world and build more realistic cloth  \n models\, and how models of the word can be used to detect forgeries that  \n otherwise fool human perception. Bio: James F. O'Brien is a Professor of  \n Computer Science at the University of California\, Berkeley. His primary area  \n of interest is Computer Animation\, with an emphasis on generating realistic  \n motion using physically based simulation and motion capture techniques. He  \n has authored numerous papers on these topics. In addition to his research  \n pursuits\, Prof. O'Brien has worked with several game companies on integrating  \n advanced simulation physics into game engines\, and his methods for  \n destruction modeling have been used in more than 15 feature films. He  \n received his doctorate from the Georgia Institute of Technology in 2000\, the  \n same year he joined the Faculty at U.C. Berkeley. Professor O'Brien is a  \n Sloan Fellow and ACM Distinguished Scientist\, Technology Review selected him  \n as one of their TR-100\, and he has been awarded research grants from the  \n Okawa and Hellman Foundations. He is currently serving as ACM SIGGRAPH  \n Director at Large. http://obrien.berkeley.edu/
END:VEVENT
BEGIN:VEVENT
UID:calendar.11746.field_date.0.308
SUMMARY:Sankaralingam Panneerselvam: System Design for Heterogeneous Architectures
DTSTAMP:20130618T164412Z
DTSTART:20130515T150000Z
DTEND:20130515T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/system-design-heterogeneous-architectures
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee:Michael Swift (advisor) Aditya Akellla David Wood Remzi  \n Arpaci-Dusseau Christopher Rossbach
END:VEVENT
BEGIN:VEVENT
UID:calendar.11729.field_date.0.309
SUMMARY:Mooly Sagiv: Concurrent Data Representation Synthesis
DTSTAMP:20130618T164412Z
DTSTART:20130515T210000Z
DTEND:20130515T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/concurrent-data-representation-synthesis
LOCATION:1240 CS
DESCRIPTION:We describe an approach for synthesizing data representations for concurrent  \n programs. Our compiler takes as input a program written using concurrent  \n relations\, and synthesizes a representation of the relations as sets of  \n cooperating data structures\, as well as the placement and acquisition of  \n locks to synchronize concurrent access to those data structures. The  \n resulting code is correct by construction: individual relational operations  \n are implemented correctly\, and the aggregate set of operations is  \n serializable and deadlock free. The relational specification also permits a  \n high-level optimizer to choose the best performing of many possible legal  \n data representations and locking strategies\, which we demonstrate with an  \n experiment autotuning a graph benchmark. This is joint work with Alex Aiken  \n and Peter Hawkins(Stanford)\, Kathleen Fisher(DARPA)\, and Martin Rinard(MIT).  \n The work is part of Peter Hawkins's Ph.D. thesis  \n (http://theory.stanford.edu/~hawkinsp/). Please also see the article about  \n the work that appeared in the December 2012 issue of CACM.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11747.field_date.0.310
SUMMARY:Vijaychidambaram Velayudhan Pillai : Optimistic Storage Systems
DTSTAMP:20130618T164412Z
DTSTART:20130516T180000Z
DTEND:20130516T200000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/optimistic-storage-systems
LOCATION:3310 Computer Sciences
DESCRIPTION:Committee: Andrea C. Arpaci-Dusseau (Advisor) Remzi H. Arpaci-Dusseau  \n (Advisor) Mark Hill Michael Swift
END:VEVENT
BEGIN:VEVENT
UID:calendar.11752.field_date.0.311
SUMMARY:Mike Rabbat\, Assistant Professor: Communication/Computation Tradeoffs in  \n Consensus-Based Distributed Optimization
DTSTAMP:20130618T164412Z
DTSTART:20130516T203000Z
DTEND:20130516T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/communicationcomputation-tradeoffs-consensus-based-distributed-optimization
LOCATION:WID room 3280 (teaching...
DESCRIPTION:We study the scalability of consensus-based distributed optimization  \n algorithms. Nodes in the network cooperate to find the global minimizer of a  \n separable convex objective function\, and each node only knows one component  \n of the objective. In consensus-based distributed optimization algorithms\,  \n nodes alternate between taking gradient descent steps based on their local  \n component of the objective\, and communicating information with neighbors to  \n reach agreement on the location of the globally optimal solution. In this  \n work we focus on two questions: How many nodes should we use to solve a given  \n problem as quickly as possible? and How often should they communicate in  \n order to reach an epsilon-optimal solution as quickly as possible? Central to  \n our analysis is a problem-specific value $r$ which quantifies the time it  \n takes to communicate one gradient relative to the time it takes to compute  \n one gradient. We show that\, in general\, when nodes communicate after every  \n gradient computation\, there is a problem-specific optimal number of  \n processors which depends on $r$ and characteristics of the network topology.  \n Surprisingly\, the time to reach a fixed level of accuracy can be reduced by  \n communicating less frequently as the computation progresses. Experiments  \n solving metric learning and non-smooth convex minimization tasks on a cluster  \n demonstrate strong agreement between theory and practice.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11739.field_date.0.312
SUMMARY:Kwang-Sung Jun: Learning from Human-Generated Lists
DTSTAMP:20130618T164412Z
DTSTART:20130523T210000Z
DTEND:20130523T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/learning-human-generated-lists
LOCATION:CS4310
DESCRIPTION:** ICML Practice Talk ** * Abstract: Human-generated lists are a form of  \n non-iid data with important applications in machine learning and cognitive  \n psychology. We propose a generative model -- sampling with reduced  \n replacement (SWIRL) -- for such lists. We discuss SWIRL's relation to  \n standard sampling paradigms\, provide the maximum likelihood estimate for  \n learning\, and demonstrate its value with two real-world applications: (i) In  \n a "feature volunteering" task where non-experts spontaneously generate  \n feature=>label pairs for text classi cation\, SWIRL improves the accuracy of  \n state-of-the-art feature-learning frameworks. (ii) In a "verbal fluency" task  \n where brain-damaged patients generate word lists when prompted with a  \n category\, SWIRL parameters align well with existing psychological theories\,  \n and our model can classify healthy people vs. patients from the lists they  \n generate.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11805.field_date.0.313
SUMMARY:Srikrishna Sridhar: Models and Algorithms for Mixed Integer Programming
DTSTAMP:20130618T164412Z
DTSTART:20130528T140000Z
DTEND:20130528T160000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/models-and-algorithms-mixed-integer-programming
LOCATION:4130 Wisconsin...
DESCRIPTION:Committee: Stephen Wright (advisor) Jeff Linderoth James Leudtke Christopher  \n Ré Tomas Rutherford
END:VEVENT
BEGIN:VEVENT
UID:calendar.11820.field_date.0.314
SUMMARY:Asim Kadav: Understanding and Improving Device Access Complexity
DTSTAMP:20130618T164412Z
DTSTART:20130604T150000Z
DTEND:20130604T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/understanding-and-improving-device-access-complexity
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: Prof. Michael M. Swift (advisor) Prof. Remzi Arpaci-Dusseau Prof.  \n Ben Liblit Prof. Somesh Jha Prof. Parmesh Ramanthan Prof. Thomas Ristenpart
END:VEVENT
BEGIN:VEVENT
UID:calendar.11824.field_date.0.315
SUMMARY:Ashish Patro: Centralized Frameworks to Understand Wireless Experience
DTSTAMP:20130618T164412Z
DTSTART:20130606T200000Z
DTEND:20130606T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/centralized-frameworks-understand-wireless-experience
LOCATION:3310 CS
DESCRIPTION:Committee: Suman Banerjee (advisor) Paul Barford Xinyu Zhang
END:VEVENT
BEGIN:VEVENT
UID:calendar.11811.field_date.0.316
SUMMARY:Jun-Ming Xu: An Examination of Regret in Bullying Tweets
DTSTAMP:20130618T164412Z
DTSTART:20130606T210000Z
DTEND:20130606T220000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/examination-regret-bullying-tweets
LOCATION:CS 4310
DESCRIPTION:Practice Talk for 2013 Conference of North American Chapter of the  \n Association for Computational Linguistics: Human Language Technologies (NAACL  \n HLT). *Abstract*: Social media users who post bullying related tweets may  \n later experience regret\, potentially causing them to delete their posts. In  \n this paper\, we construct a corpus of bullying tweets and periodically check  \n the existence of each tweet in order to infer if and when it becomes deleted.  \n We then conduct exploratory analysis in order to isolate factors associated  \n with deleted posts. Finally\, we propose the construction of a regrettable  \n posts predictor to warn users if a tweet might cause regret.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11819.field_date.0.317
SUMMARY:Danielle Albers: Perceptually Informed Scalable Sequence Comparison
DTSTAMP:20130618T164412Z
DTSTART:20130607T150000Z
DTEND:20130607T160000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/perceptually-informed-scalable-sequence-comparison
LOCATION:CS3310
DESCRIPTION:Committee: Michael Gleicher (Advisor) Steve Franconeri Kevin Ponto Bilge  \n Mutlu
END:VEVENT
BEGIN:VEVENT
UID:calendar.11825.field_date.0.318
SUMMARY:Joseph Chabarek : Toward a Power-aware Internet
DTSTAMP:20130618T164412Z
DTSTART:20130607T160000Z
DTEND:20130607T180000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/toward-power-aware-internet
LOCATION:4310 CS
DESCRIPTION:Committee: Paul Barford (advisor) Suman Banerjee Michael Swift Remzi  \n Arpaci-Dusseau Sujata Banerjee (HP-Labs)
END:VEVENT
BEGIN:VEVENT
UID:calendar.11818.field_date.0.319
SUMMARY:Michael Correll: Improving Visual Statistics for Decision-Making
DTSTAMP:20130618T164412Z
DTSTART:20130607T180000Z
DTEND:20130607T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/improving-visual-statistics-decision-making
LOCATION:CS3310
DESCRIPTION:Committee: Michael Gleicher (Advisor) Steve Franconeri Charles Franklin Bilge  \n Mutlu
END:VEVENT
BEGIN:VEVENT
UID:calendar.11821.field_date.0.320
SUMMARY:Arkaprava Basu: Efficient Virtual Memory for Big Memory Servers
DTSTAMP:20130618T164412Z
DTSTART:20130611T210000Z
DTEND:20130611T213000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/efficient-virtual-memory-big-memory-servers
LOCATION:1221 CS
DESCRIPTION:This is a practice talk for ISCA. Our analysis shows that many  \n “big-memory” server workloads\, such as databases\, in-memory caches\, and  \n graph analytics\, pay a high cost for page-based virtual memory. They consume  \n as much as 10% of execution cycles on TLB misses\, even using large pages. On  \n the other hand\, we find that these workloads use read-write permission on  \n most pages\, are provisioned not to swap\, and rarely benefit from the full  \n flexibility of page-based virtual memory. To remove the TLB miss overhead for  \n big-memory workloads\, we propose mapping part of a process’s linear virtual  \n address space with a direct segment\, while page mapping the rest of the  \n virtual address space. Direct segments use minimal hardware—base\, limit and  \n offset registers per core—to map contiguous virtual memory regions directly  \n to contiguous physical memory. They eliminate the possibility of TLB misses  \n for key data structures such as database buffer pools and in-memory key-value  \n stores. Memory mapped by a direct segment may be converted back to paging  \n when needed. We prototype direct-segment software support for x86-64 in Linux  \n and emulate direct-segment hardware. For our workloads\, direct segments  \n eliminate almost all TLB misses and reduce the execution time wasted on TLB  \n misses to less than 0.5%.
END:VEVENT
BEGIN:VEVENT
UID:calendar.11826.field_date.0.321
SUMMARY:Matthew Renzelmann : Improving the Development and Testing of Device Drivers
DTSTAMP:20130618T164412Z
DTSTART:20130612T180000Z
DTEND:20130612T200000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/improving-development-and-testing-device-drivers
LOCATION:4310
DESCRIPTION:Committee: Prof. Michael M. Swift (advisor) Prof. Remzi Arpaci-Dusseau Prof.  \n Ben Liblit Prof. Somesh Jha Prof. Parmesh Ramanathan
END:VEVENT
BEGIN:VEVENT
UID:calendar.11838.field_date.0.322
SUMMARY:Chen Zeng: On Differentially Private Mechanisms for Count-Range Queries and  \n their Applications
DTSTAMP:20130618T164412Z
DTSTART:20130613T190000Z
DTEND:20130613T190000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/differentially-private-mechanisms-count-range-queries-and-their-applications
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: Jeff Naughton (Advisor) Jin-Yi Cai AnHai Doan Christopher Re David  \n Page
END:VEVENT
BEGIN:VEVENT
UID:calendar.11839.field_date.0.323
SUMMARY:Wei Zhang: Improving Concurrent Software Reliability Via an Effect -oriented  \n Approach
DTSTAMP:20130618T164412Z
DTSTART:20130614T140000Z
DTEND:20130614T160000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/improving-concurrent-software-reliability-effect-oriented-approach
LOCATION:4310 Computer Sciences
DESCRIPTION:Committee: Shan Lu (advisor) Michael Swift Thomas Reps Karthikeyan  \n Sankaralingam Mark HIll
END:VEVENT
BEGIN:VEVENT
UID:calendar.11840.field_date.0.324
SUMMARY:C. Mohan: Global Technology Outlook (GTO) 2013
DTSTAMP:20130618T164412Z
DTSTART:20130619T160000Z
DTEND:20130619T170000Z
URL;VALUE=URI:http://www.cs.wisc.edu/event/global-technology-outlook-gto-2013
LOCATION:CS 4310
DESCRIPTION:AbstractThe Global Technology Outlook (GTO) is IBM Research's vision of the  \n future for information technology (IT) and its impact on industries that use  \n IT. This annual exercise highlights emerging software\, hardware\, and services  \n technology trends that are expected to significantly impact the IT sector in  \n the next 3-7 years. In particular\, the GTO identifies technologies that may  \n be disruptive to an existing business\, have the potential to create new  \n opportunities\, and can provide new business value to our customers. The 2013  \n GTO is built not only on its 31 predecessors\, but the 100 years of IBM  \n innovation. The 2013 GTO reports on six key findings which form 2 groups. The  \n first group addresses The Rapidly Evolving Infrastructure while the second  \n one addresses The Future of Big Data and Analytics. The six topics of GTO  \n 2013 are: Mobile First\, Scalable Services Ecosystems\, Software Defined  \n Environments\, Multimedia and Visual Analytics\, Contextual Enterprise and  \n Personalized Education. In this talk\, I will share the GTO 2013 findings with  \n the audience. This talk should be of interest to not only technical people  \n but also to a much broader set of people.BiographyDr. C. Mohan has been an  \n IBM researcher for 30 years in the information management area\, impacting  \n numerous IBM and non-IBM products\, the research community and standards\,  \n especially with his invention of the ARIES family of locking and recovery  \n algorithms\, and the Presumed Abort commit protocol. This IBM\, ACM and IEEE  \n Fellow has also served as the IBM India Chief Scientist. In addition to  \n receiving the ACM SIGMOD Innovation Award\, the VLDB 10 Year Best Paper Award  \n and numerous IBM awards\, he has been elected to the US and Indian National  \n Academies of Engineering\, and has been named an IBM Master Inventor. This  \n distinguished alumnus of IIT Madras received his PhD at the University of  \n Texas at Austin. He is an inventor of 38 patents. He serves on the advisory  \n board of IEEE Spectrum and on the IBM Software Group Architecture Board’s  \n Council. More information can be found in his home page at bit.ly/CMohan
END:VEVENT
END:VCALENDAR