Title

Efficient learning algorithm for associative memory neural network

Document Type

Article

Publication Date

12-1-1997

Abstract

A new efficient learning algorithm of associative memory neural network is proposed, with the following characteristics: (1) it can store any given training pattern set no matter how much and what correlation among them may be; (2) the smallest domain of attraction of training patterns is maximized; (3) each domain of attraction of training patterns is guaranteed to be as large as possible; (4) the designed associative memory network is globally stable. A large number of computer experimental results confirm that this algorithm possesses more powerful storage ability and more fault-tolerance capability than existing ones.

Publication Source (Journal or Book title)

Zidonghua Xuebao/Acta Automatica Sinica

First Page

721

Last Page

727

This document is currently not available here.

COinS