NIPS Proceedingsβ

Smoothness, Low Noise and Fast Rates

Part of: Advances in Neural Information Processing Systems 23 (NIPS 2010)

[PDF] [BibTeX] [Supplemental]

Authors

Abstract

We establish an excess risk bound of O(H R_n^2 + sqrt{H L*} R_n) for ERM with an H-smooth loss function and a hypothesis class with Rademacher complexity R_n, where L* is the best risk achievable by the hypothesis class. For typical hypothesis classes where R_n = sqrt{R/n}, this translates to a learning rate of ̃ O(RH/n) in the separable (L* = 0) case and O(RH/n + sqrt{L* RH/n}) more generally. We also provide similar guarantees for online and stochastic convex optimization of a smooth non-negative objective.