Elsevier

Journal of Complexity

Volume 24, Issue 3, June 2008, Pages 380-397
Journal of Complexity

Obtaining fast error rates in nonconvex situations

https://doi.org/10.1016/j.jco.2007.09.001Get rights and content
Under an Elsevier user license
open archive

Abstract

We show that under mild assumptions on the learning problem, one can obtain a fast error rate for every reasonable fixed target function even if the base class is not convex. To that end, we show that in such cases the excess loss class satisfies a Bernstein type condition.

Keywords

Nearest point map
Nonconvex classes
Bernstein condition

Cited by (0)

Supported in part by Australian Research Council discovery Grant DP0559465 and by Israel Science Foundation Grant ISF 666/06.