Olvi Mangasarian reflects on pioneering career in optimization

“Nothing happens in the universe that does not have a sense of either certain maximum or minimum” reads a favorite quote of Professor Emeritus Olvi Mangasarian.  It comes from an eighteenth-century Swiss mathematician, Leonhard Euler.

Mangasarian, who joined the University of Wisconsin-Madison faculty in the late '60s, loves the elegance and versatility of the field of optimization.  He’s used its principles to tackle problems ranging from cancer treatment to the authorship of historical documents.

His advice for today’s students?  “No matter what area you’re in, get at least a taste of optimization.  You don’t know where in your field of choice you can find an application that will get new results.”

Mangasarian’s colleague (and former student) Professor Michael Ferris, who has known him since 1985, names him one of the greats in their field.  In 2011, Mangasarian became a fellow of the Society for Industrial and Applied Mathematics (SIAM) for his seminal contributions to the theory and algorithms of optimization and applications to machine learning.

Says Ferris, who earned his PhD at Cambridge University but spent some of his graduate years in Madison as Mangasarian’s research assistant, “He’s one of the main reasons I came here to study.  He’s a leader in nonlinear programming and, subsequently, he generated a new field of optimization applied to machine learning and pattern recognition.  He was at the forefront of the development of that area, now called data analytics,” having first published on it in the '60s.

Just as important, Ferris says, is Mangasarian’s collaborative spirit.  “He’s extremely focused and very inclusive.  He shares ideas and pushes and guides people in the right research direction.  He’s a great mentor.”  

 

     Mangasarian earned his PhD at Harvard University in 1959—before any American university had even established a computer science department.  His focus at Harvard was applied mathematics, following bachelor’s and master’s degrees in engineering at Princeton.

After spending eight years working at a Shell Oil subsidiary, Mangasarian came to Madison in 1967, when the computer science department was still in its infancy.  The late Professor Ben Rosen, who had been Mangasarian’s boss at Shell, was by then the third department chair.  Rosen helped lure Mangasarian from private industry to academia.

One of Mangasarian’s most notable achievements occurred in collaboration with William Wolberg, professor emeritus in the UW-Madison School of Medicine and Public Health.  Their chance meeting at a party led to a project in which the two applied linear-programming-based machine learning approaches to the diagnosis and prognosis of patients with breast cancer.

The duo’s work allowed doctors to extract much smaller tissue samples from patients—collected via fine-needle aspiration rather than more invasive methods—and still get accurate results as to whether tissue was benign or malignant, as well as the patient’s overall outlook.

In a much different arena, Mangasarian, together with his graduate student, Glenn Fung (MS '01, PhD '03), clarified the authorship of the twelve disputed Federalist Papers.  (Authorship was clear for several dozen papers published by James Madison, Alexander Hamilton or John Jay in 1787-88, but a dozen more were in question.)

Statisticians had already approached the problem with methods based on the frequency of about 70 words used in the essays.  Fung and Mangasarian employed a simpler solution that gave more visually clear-cut results.  Using feature extraction, they separated the papers of Madison and Hamilton using only three words (“would,” “to,” and “upon”) and plotting them in three dimensions.  Confirming the statisticians’ study, the UW study found that the disputed papers were indeed James Madison’s.

Mangasarian, who retired in 2003, looks back with pride on the 27 doctoral students whose work he supervised at UW-Madison from the 1970s to the 2000s.  His career has spanned over 200 published papers and several books, as well as helping to establish a Data Mining Institute with his colleagues Michael Ferris, Jeff Naughton, Raghu Ramakrishnan (now at Microsoft) and the late Jim Gray.

While he travels less frequently these days, Mangasarian and his wife enjoy summers in Madison and winters in San Diego, where he has held a research appointment since 2002 at the University of California, San Diego.

And though classical music occupies some of his free time, research in mathematical programming, machine learning and data mining still consumes his energy.  “The challenges are still there, and I may be able to contribute, to keep my mind active,” he says.

 

[Credit for top photo: Olvi Mangasarian and Michael Ferris in Spoleto, Italy, in 1989.  Photo courtesy of Michael Ferris]