SYNAPTIC WEIGHT NOISE DURING MULTILAYER PERCEPTRON TRAINING - FAULT-TOLERANCE AND TRAINING IMPROVEMENTS

Authors
MURRAY AF EDWARDS PJ
Citation
Af. Murray et Pj. Edwards, SYNAPTIC WEIGHT NOISE DURING MULTILAYER PERCEPTRON TRAINING - FAULT-TOLERANCE AND TRAINING IMPROVEMENTS, IEEE transactions on neural networks, 4(4), 1993, pp. 722-725
Citations number
8
Categorie Soggetti
Computer Application, Chemistry & Engineering","Engineering, Eletrical & Electronic","Computer Applications & Cybernetics
Journal title
IEEE transactions on neural networks → ACNP
ISSN journal
10459227
Volume
4
Issue
4
Year of publication
1993
Pages
722 - 725
Database
ISI
SICI code
1045-9227(1993)4:4<722:SWNDMP>2.0.ZU;2-V
Abstract
We develop a mathematical model of the effects of synaptic arithmetic noise in multilayer perceptron training. Predictions are made regardin g enhanced fault-tolerance and generalization ability, and also an imp roved learning trajectory. These predictions are subsequently verified by simulation. The results are perfectly general and have profound im plications for the accuracy requirements in MLP training, particularly in the analog domain.