联想Hopfield神经网络样本稳定性的判别方法  

New distinguishing method of prototype stability about associative Hopfield neural networks

作　　者：赵婷婷[1] 魏小鹏[2] 陈刚[3] 

机构地区：[1]大连理工大学机械工程系,辽宁大连116023 [2]大连大学先进设计技术中心,辽宁大连116622 [3]辽宁工程技术大学基础部,辽宁阜新123000

出　　处：《辽宁工程技术大学学报（自然科学版）》2004年第1期135-137,共3页Journal of Liaoning Technical University (Natural Science)

基　　金：国家自然科学基金资助(60174037;50275013)

摘　　要：针对基于Hebb学习的离散Hopfield神经网络,论述了实现联想记忆功能的基本方法及其样本稳定性的基本思想。根据所存在的问题,提出了判别样本稳定性的充要条件,以及判别样本稳定的一个充分条件,使判别样本是否稳定变得简单易行。最后给出了判别样本稳定的矩阵表示形式,并对给出的定理以及推论给予了证明。In view of discrete Hopfield neural networks based on Hebb learning law, the essential method of achieving association and memory function and the essential thinking of prototype stability are discussed. According to the existing problem, the necessary and sufficient conditions of distinguishing prototype stability were put forward. A sufficient condition of distinguishing prototype stability was given. This makes it very easy to distinguish prototype stability. At last, the matrix form of distinguishing prototype stability was given. All theorems and deductions have been proved.

关 键 词：HOPFIELD神经网络 样本稳定性 HEBB学习规则 联想记忆 矩阵 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]