Please use this identifier to cite or link to this item:
Title: Error tolerant environment of multilayer perceptrons with controlled learning
Authors: Yu, Wellington C.P.
Teh, Hoon-Heng 
Issue Date: 1988
Citation: Yu, Wellington C.P.,Teh, Hoon-Heng (1988). Error tolerant environment of multilayer perceptrons with controlled learning. Neural Networks 1 (1 SUPPL) : 323-. ScholarBank@NUS Repository.
Abstract: The back-propagation learning algorithm of multilayer perceptrons has provided a way to determine the weight-matrices. However, besides giving a learned solution, the method does not reveal the structure of the multilayer perceptrons, the number of layers, and the number of nodes, for performing further analysis. To solve the shortcomings of back-propagation learning, we have developed an algorithm to obtain quickly an exact upper-bound solution for a 2-layer perceptron for a given n input patterns of size m, and n output patterns of size k. For reduction of nodes in the hidden unit, we have developed an algorithm to minimize the hidden layer structure for an exact lower-bound solution. To cope with the problem of not error tolerant, we have developed an algorithm to modify the weight and threshold functions so that inputs with errors can be tolerated.
Source Title: Neural Networks
ISSN: 08936080
DOI: 10.1016/0893-6080(88)90354-1
Appears in Collections:Staff Publications

Show full item record
Files in This Item:
There are no files associated with this item.

Page view(s)

checked on Oct 10, 2020

Google ScholarTM



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.