摘要
用最小二乘法产生的神经元联系强度矩阵取代 Hopfield 模型中的矩阵后,增强了网络的识别能力,新模型与其它几个模型的实验比较也在文章后面给出.
It is shown that modifying the Hopfield neural network model by using LMS rule rather than other rules increases the recognition capability of the network. An cxperimental compar.son with other methods is presented.
关键词
模式识别
神经元网络
记忆矩阵
neural networks
connection matrx
paternocognition
