Title :
An algorithm for training multilayer perceptrons
Author_Institution :
Texas A&M Univ., College Station, TX
Abstract :
Summary form only given, as follows. An algorithm for training multilayer perceptrons with hard-threshold functions was proposed. The proposed algorithm is guaranteed to classify correctly any given set of patterns, and therefore alleviates some of the drawbacks of the back-propagation algorithm, such as the frequent failure to converge to the global minimum. The network considered is a two-layer network (one hidden layer), and the algorithm is of the incremental type, which means that neurons continue to be added in some way until the training patterns are correctly classified
Keywords :
learning systems; neural nets; hard-threshold functions; multilayer perceptrons; neural network; training; two-layer network; Backpropagation algorithms; Multilayer perceptrons; Neurons;
Conference_Titel :
Neural Networks, 1991., IJCNN-91-Seattle International Joint Conference on
Conference_Location :
Seattle, WA
Print_ISBN :
0-7803-0164-1
DOI :
10.1109/IJCNN.1991.155629