Feature Selection Using a Multilayer Perceptron

Published 1 January 1990

Dennis W. Ruck, Steven K. Rogers, Matthew Kabrisky

Citations226

TL;DR

A technique has been developed which analyzes the weights in a multilayer perceptron to determine which features the network finds important and which are unimportant, and the saliency measure is used to compare the results of two different training rules on a target recognition problem.

Abstract

The problem of selecting the best set of features for target recognition using a multilayer perceptron is addressed in this paper. A technique has been developed which analyzes the weights in a multilayer perceptron to determine which features the network finds important and which are unimportant. A brief introduction to the use of multilayer perceptrons for classification and the training rules available is followed by the mathematical development of the saliency measure for multilayer perceptrons. The technique is applied to two different image databases and is found to be consistent with statistical techniques and independent of the network initial conditions. The saliency measure is then used to compare the results of two different training rules on a target recognition problem. 1 Introduction Recently there has been a great deal of interest in the use of multilayer perceptrons as classifiers in pattern recognition problems (see, for example, [11]). Unfortunately, little ...

Keywords

Arts and HumanitiesHealth Professions

Mathematics of Control Signals and SystemsApproximation by superpositions of a sigmoidal function

13,684 Citations1989George Cybenko

Pattern classification and scene analysis

12,643 Citations1973Richard O. Duda, Peter E. Hart

IEEE Transactions on ComputersAircraft Identification by Moment Invariants

754 Citations1977Sahibsingh A. Dudani, Kenneth J. Breeding +1 more

An experimental system is described in which certain features called moment invariants are extracted from binary television images and are used for automatic classification of aircraft types from optical images, exhibiting a significantly lower error rate than human observers.

Neural Information Processing SystemsTraining Multilayer Perceptrons with the Extended Kalman Algorithm

319 Citations1988Sharad Singhal, Lance Wu

It is shown that training multilayer perceptrons is an identification problem for a nonlinear dynamic system which can be solved using the Extended Kalman Algorithm.

IEEE Transactions on Information TheoryConsiderations of sample and feature size

292 Citations1972Donald H. Foley

The design-set error rate for a two-class problem with multivariate normal distributions is derived as a function of the sample size per class (N) and dimensionality (L) and is demonstrated to be an extremely biased estimate of either the Bayes or test- set error rate.

Statistical properties of artificial neural networks

80 Citations2003Andrew R. Barron

It is shown how, with the use of an appropriate complexity regularization criterion, the statistical risk of network estimators converges to zero as the sample size increases.

Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIEAn Approach To Multiple Sensor Target Detection

9 Citations1989Michael C. Roggemann, James P. Mills +3 more

Two single sensor and three multiple sensor approaches to target detection were developed, and their performance compared, and performance improvements are described which resulted from incorporating the correspondence feature information into the class estimation process.