Local Rademacher complexities
Statistics Theory
2007-06-13 v1 Statistics Theory
Abstract
We propose new bounds on the error of learning algorithms in terms of a data-dependent notion of complexity. The estimates we establish give optimal rates and are based on a local and empirical version of Rademacher averages, in the sense that the Rademacher averages are computed from the data, on a subset of functions with small empirical error. We present some applications to classification and prediction with convex function classes, and with kernel classes in particular.
Cite
@article{arxiv.math/0508275,
title = {Local Rademacher complexities},
author = {Peter L. Bartlett and Olivier Bousquet and Shahar Mendelson},
journal= {arXiv preprint arXiv:math/0508275},
year = {2007}
}
Comments
Published at http://dx.doi.org/10.1214/009053605000000282 in the Annals of Statistics (http://www.imstat.org/aos/) by the Institute of Mathematical Statistics (http://www.imstat.org)