基于广义逆的指数型联想存储器模型

EXPONENTIAL ASSOCIATIVE MEMORY MODELBASED ON GENERALIZED INVERSE

基于Ｋｏｈｏｎｅｎ的广义逆联想存储模型ＧＩＡＭ（ｇｅｎｅｒａｌｉｚｅｄｉｎｖｅｒｓｅａｓｏｃｉａｔｉｖｅｍｅｍｏｒｙ）和Ｍｕｒａｋａｍｉ的最小平方联想存储ＬＳＡＭ（ｌｅａｓｔｓｑｕａｒｅｓａｓｓｏｃｉａｔｉｖｅｍｅｍｏｒｙ）原理，本文提出了一个指数型联想存储器．该模型的存储性能经计算机模拟证实，远远优于ＧＩＡＭ和ＬＳＡＭ，通过适当地调节参数，几乎可达到完全的联想．对输入噪声方差，无需先验假设。

Based on Kohonen's GIAM(generalized inverse associative memory) and Murakami's LSAM(least squares associative memory) principles, an exponential associative memory is presented in this paper. The computer simulations have shown that the associative performance of the proposed model is superior to those of GIAM and LSAM, and its recall for stored data is almost perfect only via adjusting its parameter. The model does not require a prior assumption to noise variances and realizes nonlinear mapping between inputs and outputs to some extent.