双向联想记忆神经网络的一种编码策略


Autoria(s): 于海斌; 薛劲松; 王浩波; 徐心和
Data(s)

1997

Resumo

本文提出一种双向联想记忆神经网络的按‘位’加权编码策略,并给出了求取权值的速推算法.它将Kosko双向联想记忆神经网络按海明距离进行模式匹配的原则,修正为按加权海明距离进行模式匹配,从而可以使得对不满足连续性的所谓“病态结构”的一类样本模式集,同样具有良好的联想能力.对二值图象模式存贮、联想的计算机模拟实验表明,此方法具有优良的性能和实用价值。

A coding strategy by weighted bits and a recursive algorithm of obtaining theweights for bidirectional associative memories are presented in the paper. This strategy modifiesKosko's bidirectional associative memories matching patterns by Hamming distance into those byweighted Hamming distance so that ill structured sample pattern set without continuity can also bestored well. It is manifested that this method has high performance and practicability through storing and associating binary pixel patterns with computer simulations.

国家高技术863计划资助

Identificador

http://ir.sia.ac.cn//handle/173321/1345

http://www.irgrid.ac.cn/handle/1471x/172385

Idioma(s)

中文

Palavras-Chave #神经网络 #双向联想记忆 #海明距离 #最速下降法
Tipo

期刊论文