近存储饱和状态下联想学习记忆的神经网络模型

AN ASSOCIATION-LEARNING-MEMORY NEURAL NETWORK MODEL NEAR SATURATION STATE OF STORAGE

本文提出了神经网络在近饱和状态下的一种联想学习记忆模型.讨论了该模型的主要特性,对由100个神经元、记忆10个随机图样组成的网络系统给出并分析了计算机模拟结果,讨论了该模型的学习律与传统的Hebb学习律的区别,研究了网络在学习记忆和联想新态时初始噪声Pi和联想噪声Pa对新态恢复行为的影响,总结了在近饱和状态下该模型所具有的优势.

This paper Presents a new neural network model near saturation state of storage and discusses main properties of the model on association and learning.Calculations of computer simulation to the system with 100 neurons and 10 random patterns are given and analysed. Discussion of the difference between the new model and Hebb's rule is presented, and the effects of initial noise Pi and association noise P. on the retrival of a learned pattern are also studied.Some conclusions are obtained.