An invariance property of kernel based predictors

Thumbnail Image

Date

Authors

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Description

We consider kernel based learning methods for regression and analyze what happens to the risk minimizer when new variables, statistically independent of input and target variables, are added to the set of input variables; this problem arises, for example, in the detection of causality relations between two time series. We find that the risk minimizer remains unchanged if we constrain the risk minimization to hypothesis spaces induced by suitable kernel functions. We show that not all kernel induced hypothesis spaces enjoy this property. We present sufficient conditions ensuring that the risk minimizer does not change, and show that they hold for inhomogeneous polynomial and Gaussian RBF kernels. We also provide examples of kernel induced hypothesis spaces whose risk minimizer changes if independent variables are added as input.
Revised version in press in Neural Computation

Keywords

Disordered Systems and Neural Networks, Data Analysis, Statistics and Probability

Citation

URI

http://salesiana.dossiersoluciones.com/handle/123456789/100424

Consulte el texto completo en el siguiente enlace:

https://arxiv.org/abs/cond-mat/0502511
 http://arxiv.org/abs/cond-mat/0502511

Collections

arXiv