Configuring interaction of memorized patterns with an asymmetric Hebbian rule for recurrent neural networks

We propose an asymmetric Hebbian rule for connection weights of recurrent associative neural networks. By using this model, networks perform several new properties such as simultaneously embedding patterns of various activities, distinguishing memorized patterns from others and functional associatio...

Full description

Saved in:
Bibliographic Details
Published in:Neurocomputing (Amsterdam) 1996, Vol.10 (1), p.43-53
Main Authors: Ishii, Toshinao, Kyuma, Kazuo
Format: Article
Language:English
Subjects:
Associative memory
Asymmetric connection
Cooperative association
Distributed representation
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We propose an asymmetric Hebbian rule for connection weights of recurrent associative neural networks. By using this model, networks perform several new properties such as simultaneously embedding patterns of various activities, distinguishing memorized patterns from others and functional association by configuring cooperations of memories within the associative dynamics. The asymmetric Hebbian model is derived from a requirement of distributed information representation. Specifically, connection weights are defined so that the neural dynamics do not depend on the activity of the inputs.
ISSN:0925-2312
1872-8286
DOI:10.1016/0925-2312(95)00046-1