离散型Hopfield神经网络稳定性的研究

On stability of time Hopfield neural network

提出了离散型Hopfield神经网络DHNN模型权矩阵的等效矩阵的概念,给出了相关算法。利用等效矩阵的概念可把DHNN模型按其轨道是否相同进行分类。因而,获得了研究DHNN模型收敛性的一种新方法,并利用此方法推广了已有的有关收敛性的几乎所有结论。

A new concept on the equivalent matrices of discrete Hopfield neural network (DHNNS) is introduced. A related algorithm on equivalent matrices is presented. Based on the new concept, a classification of DHNNS with neurons is established in such a way that two DHNNS belonging to the same class have completely the same dynamic property. A new way to study the stablility about DHNNS is siven and almost all of results on the stability about DHNNS are expanded from this.