Some Theoretical Results Concerning the Convergence of Compositions of Regularized Linear Functions

Part of Advances in Neural Information Processing Systems 12 (NIPS 1999)

Bibtex Metadata Paper

Authors

Tong Zhang

Abstract

Recently, sample complexity bounds have been derived for problems in(cid:173) volving linear functions such as neural networks and support vector ma(cid:173) chines. In this paper, we extend some theoretical results in this area by deriving dimensional independent covering number bounds for regular(cid:173) ized linear functions under certain regularization conditions. We show that such bounds lead to a class of new methods for training linear clas(cid:173) sifiers with similar theoretical advantages of the support vector machine. Furthermore, we also present a theoretical analysis for these new meth(cid:173) ods from the asymptotic statistical point of view. This technique provides better description for large sample behaviors of these algorithms.