ICML-98 Submission #58

Ridge Regression learning algorithm in Dual Variables

Authors:

C. Saunders, A. Gammerman, V. Vovk
Royal Holloway and New Bedford College,
University of London,
Egham,
Surrey,
TW20 0EX.

Abstract:

In this paper we present a dual version for Ridge Regression.  
It allows us to perform non-linear regression by constructing a linear
regression function in a high dimensional feature space.  The feature
space representation can result in a large increase in the number of
parameters used by the algorithm.  In order to combat this ``curse of
dimensionality'', the algorithm allows the use of kernel functions, as
used in Support Vector methods.  A new family of kernel functions
constructed using ANOVA decomposition are also discussed.  This paper
introduces a new regression estimation algorithm which is a
combination of these two techniques. Experimental results are then
presented (based on the Boston Housing data) which indicate the
performance of dual-variable Ridge Regression relative to other
algorithms.

Keywords : Support Vector, ANOVA decomposition

Contact author email : A.Gammerman@dcs.rhbnc.ac.uk

Contact author phone : +44 1784 443434