Asymptotitically [sic] stable recurrent neural networks : theory and application
Date
1997
Authors
Dorocicz, John Tadeusz
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This thesis will outline the basis for some types of artificial neural networks. In particular a special type of recurrent neural network that has been proven to be asymptotically stable will be studied. The theory required to train such a neural network will be presented. A number of heuristics designed to improve the training of the asymptotically stable recurĀrent neural network and the results of these heuristics on a variety of data sets will be preĀsented. An application for the asymptotically stable recurrent neural network involving the monitoring of cable television trunk amplifiers will also be described.