报告问题：Variants of the Support Vector Machine and Their Applications to Microarray Classification

报告人：Dr. ZHU Ji, Assistant Professor,

Department of Statistics, University of Michigan

时 间：2005.7.4下昼2：00

地 点：伟易博116室

摘要：The support vector machine is a widely used tool for classification. In this talk, we start with a brief introduction to the standard 2-norm support vector machine, and write it as a regularized optimization problem. Based on that, we consider several variants of the support vector machine, specifically, penalized logistic regression, the 1-norm support vector machine, and the doubly regularized support vector machine. We argue that these variants may have some advantages over the standard 2-norm support vector machine under certain situations, for example, when there are redundant noise variables. We also propose efficient algorithms to solve the optimization problems posed by these variants. In the end, we compare these models on a microarray cancer dataset.

This talk consists of a collection of joint work with Saharon Rosset

(IBM), Hui Zou (Stanford), Trevor Hastie (Stanford) and Rob Tibshirani

(Stanford).