双向联想记忆学习与回忆的研究

本文对双向联想记忆(BAM)的学习与回忆过程进行了详细的分析。在学习过程中,先是运用自适应非对称BAM算法进行学习,进而采用设置印象门限的反复记忆算法进行学习,本文从理论上证明了印象门限与样本吸引域之间的关系,指出反复记忆方法的理论依据。回忆过程中,采用非零阈值函数的运行方程,提出了阈值学习方法,并且从理论上证明了非零阈值函数的运行方程的采用,可进一步扩大吸引域。为了进一步扩大网络的信息存储量,本文引入了并联的BAM结构。本文方法的采纳,使得BAM网络的信息存储量、误差校正能力等得到很大程度的提高。

ON the Learing and Recalling of Bidirectional Associative Memory

This paper mainly discusses the opattern recognition problem with the method of bidirectional associative memory(BAM) neurl network. The learning and recalling process of BAM is analysed in detail. In the process of leatning, the algoithm of adaptive asymmetric AA-BAM is used. Furthermore, the algorithm of repeated memory with attraction is proved, which presents the theoretical basis of the repeated memory algorithm. In the process of recalling, the non-zero threshold function is used as the running function, an algorithm of threshold training is proposed, and a theorem is proved that the region of attraction can be widened with the use of non-zero threshold function as the running function. To increase the storage capacity of this network, the parallel structure is taken, with all of the methods in this paper, both the storage capacity and the error-correcting capability of the BAM neural network will be improved greatly.