Personal tools
You are here: Home Publications Bounds on error expectation for SVM
Document Actions

V. Vapnik and O. Chapelle (2000)

Bounds on error expectation for SVM

In: Advances in Large Margin Classifiers, ed. by A.J. Smola and P.L. Bartlett and B. Schölkopf and D. Schuurmans, pp. 261-280, Cambridge, MA, MIT Press.

Bounds on the Error Expectation for SVM in terms of the leave-one-out estimate and the expected value of certain properties of the SVM are given. It is shown that previous bounds involving the minimum margin and the diameter $D$ of the set of support vectors can be improved by the replacement of $D^2$ by $SD$. Here, $S$ is a new geometric property of the support vectors called the span. Experimental results show that this improvement gives significantly better predictions of test error than the previous bounds, and seems likely to be useful for model selection.

by admin last modified 2007-01-31 11:07

Powered by Plone CMS, the Open Source Content Management System