Training multilayer perceptron classifiers based on a modified support vector method

Author

Suykens, J.A.K. ; Vandewalle, J.

Author_Institution

Dept. of Electr. Eng., Katholieke Univ., Leuven, Belgium

Volume

10

Issue

4

fYear

1999

fDate

7/1/1999 12:00:00 AM

Firstpage

907

Lastpage

911

Abstract

In this paper we describe a training method for one hidden layer multilayer perceptron classifier which is based on the idea of support vector machines (SVM). An upper bound on the Vapnik-Chervonenkis (VC) dimension is iteratively minimized over the interconnection matrix of the hidden layer and its bias vector. The output weights are determined according to the support vector method, but without making use of the classifier form which is related to Mercer´s condition. The method is illustrated on a two-spiral classification problem

Keywords

iterative methods; learning (artificial intelligence); minimax techniques; multilayer perceptrons; pattern classification; SVM; VC dimension; Vapnik-Chervonenkis dimension; bias vector; interconnection matrix; iterative minimization; modified support vector method; one-hidden-layer multilayer perceptron classifier training; support vector machines; two-spiral classification problem; upper bound; Backpropagation; Kernel; Multilayer perceptrons; Quadratic programming; Radial basis function networks; Risk management; Support vector machine classification; Support vector machines; Upper bound; Virtual colonoscopy;

fLanguage

English

Journal_Title

Neural Networks, IEEE Transactions on

Publisher

ieee

ISSN

1045-9227

Type

jour

DOI

10.1109/72.774254

Filename

774254