多模式对连接权矩阵的一种神经网络学习算法  

A KIND OF NEURAL NETWORKS LEARNING ALGORITHMS FOR MULTIPLE-PATTERN PAIRS WEIGHTED MATRIX

作　　者：杨群生[1] 余英林[1] 

机构地区：[1]华南理工大学电子与通信工程系,广州510641

出　　处：《电子与信息学报》2001年第3期280-285,共6页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(69772026);广东省自然科学基金

摘　　要：该文提出多模式对连接权矩阵的一种神经网络学习算法,并给出了严格的理论证明。该算法能够将多个模糊模式对可靠地编码存储到尽可能少的连接权矩阵中,从而大大地减少存储空间,而且容易实现,并举例验证了它的有效性。A kind of neural networks learning algorithms for multiple-pattern pairs weighted matrix of fuzzy associative memories(FAMs) and its strict theoretic proofs are presented in this paper. Multiple fuzzy pattern pairs can be encoded to store in FAM connection weight matrixes as few as possible by the algorithm, so it can cut down storage space greatly and this algorithm can easily be implimented. Its effectiveness is testified by an example.

关 键 词：模糊联想记忆 连接权矩阵 神经网络 多模式 学习算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]