Theoretical investigation of the robustness of multilayer perceptrons: analysis of the linear case and extension to nonlinear networks

Author

Kerlirzin, Philippe ; Réfrégier, Philippe

Author_Institution

Central Res. Lab., Thomson-CSF, Orsay, France

Volume

6

Issue

3

fYear

1995

fDate

5/1/1995 12:00:00 AM

Firstpage

560

Lastpage

571

Abstract

In this paper, we address the problem of robustness in multilayer perceptrons. We present the main theoretical results in the case of linear neural networks with one hidden layer in order to go beyond the empirical study we previously made. We show that the robustness can greatly be improved and that even without decreasing performance in normal use. Finally, we show how this behavior, clearly demonstrated in the linear case, is an approximation of the behavior of nonlinear networks

Keywords

backpropagation; circuit stability; data compression; multilayer perceptrons; optimisation; performance evaluation; approximation; backpropagation; hidden layer; linear neural networks; multilayer perceptrons; nonlinear neural networks; principal component analysis; robustness; Backpropagation algorithms; Computer aided software engineering; Constraint optimization; Degradation; Image coding; Multilayer perceptrons; Neural networks; Nonhomogeneous media; Performance loss; Robustness;

fLanguage

English

Journal_Title

Neural Networks, IEEE Transactions on

Publisher

ieee

ISSN

1045-9227

Type

jour

DOI

10.1109/72.377963

Filename

377963