Title

Model Selection For Classification With A Large Number Of Classes

Keywords

Classification; High dimensional data; Low sample size; Multivariate analysis

Abstract

In the present paper, we study the problem of model selection for classification of high-dimensional vectors into a large number of classes. The objective is to construct a model selection procedure and study its asymptotic properties when both, the number of features and the number of classes, are large. Although the problem has been investigated by many authors, we research a more difficult version of a less explored random effect model where, moreover, features are sparse and have only moderate strength. The paper formulates necessary and sufficient conditions for separability of features into the informative and noninformative sets. In particular, the surprising conclusion of the paper is that separation of features becomes easier as the number of classes grows.

Publication Date

1-1-2014

Publication Title

Springer Proceedings in Mathematics and Statistics

Volume

74

Number of Pages

251-257

Document Type

Article; Proceedings Paper

Personal Identifier

scopus

DOI Link

https://doi.org/10.1007/978-1-4939-0569-0_23

Socpus ID

84920058301 (Scopus)

Source API URL

https://api.elsevier.com/content/abstract/scopus_id/84920058301

This document is currently not available here.

Share

COinS