We introduce and study several measures of complexity of functions from the convex hull of a given base class. These complexity measures take into account the sparsity of the weights of a convex combination as well as certain clustering properties of the base functions involved in it. We prove new upper confidence bounds on the generalization error of ensemble (voting) classification algorithms that utilize the new complexity measures along with the empirical distributions of classification margins, providing a better explanation of generalization performance of large margin classification methods.
@article{1123250220,
author = {Koltchinskii, Vladimir and Panchenko, Dmitry},
title = {Complexities of convex combinations and bounding the generalization error in classification},
journal = {Ann. Statist.},
volume = {33},
number = {1},
year = {2005},
pages = { 1455-1496},
language = {en},
url = {http://dml.mathdoc.fr/item/1123250220}
}
Koltchinskii, Vladimir; Panchenko, Dmitry. Complexities of convex combinations and bounding the generalization error in classification. Ann. Statist., Tome 33 (2005) no. 1, pp. 1455-1496. http://gdmltest.u-ga.fr/item/1123250220/