具有认知判别功能的自联想记忆神经网络模型

AN AUTOASSOCIATIVE MEMORY NEURAL NETWORK MODEL WITH COGNITIVE JUDGEMENT FUNCTION

提出一种具有认知判别功能的自联想记忆神经网络模型．对一个新的输入模式，当网络达到平衡点时，通过一种判别函数可判别该模式是否是网络中已存贮的模式．如不是则对其进行学习；如是则该平衡状态即为其联想结果，给出了模型及其学习规则，分析了网络吸引子半径与有关参数的关系及模型对输入模式的筛选作用，最后给出了计算结果．

This paper describes an autoassociative memory neural network model with a cognitive judgement function. For a new input pattern the model can judge whether it is a stored one through a judgement function. When the network reaches equilibrium, the equilibrium is a result of association, if the network does not reach equilibrium, it starts to learn the pattern. Thus,the fact that the network can associste and learn shows that the model is closer to human brain.This paper presents the model and its learning rule. It also analyzes the relation between the radius of the domains of attractivity of the attractor in the network and the relative parameter of themodel. It also eXplains the sifting function of the model for the input patterns. Finally, the simulation results are given.