THE HISTORY OF THE SUPPORT VECTOR METHOD
Professor Alexei Chervonenkis, Institute of Control Sciences, Moscow

Abstract: This talk will review developments over the last 30 years in the field of statistical learning theory. In particular, the following topics will be addressed:

1. "Generalized portrait" as a minimax vector that characterizes a class in pattern recognition cases.
2. Converting the problem of "generalized portrait" search to a convex programming problem.
3. Support vectors and their properties implied by the Kuhn-Tucker theorem.
4. Evaluation of generalization performance using the number of support vectors.

This seminar was held at the Department of Computer Science, Royal Holloway, University of London on 25 March 1998.

back