摘要
提出了最大最小联想记忆网络的一种动态调整学习算法，分析了动态调整算法所设计出的连接权阵网络对记忆模式对吸引域的影响．在一定条件下，它能够简便而有效地对训练模式进行联想推理．首先给出了一种快速调整学习算法，再进一步发展了一个动态指数细调规则学习算法，它以快速调整学习算法的结果作为连接权矩阵的迭代初值．计算机实验结果表明了所提学习算法的优越性．
In this paper, a dynamical learning algorithm for Fuzzy associative memorynetworks is proposed. The relation between the basin of attraction and theadjusting parameter in the networks is analysized. With a certain condition, thenew algorithm can be applied to associative inference. In the first, a quickdynamical adjusting algorithm is given, and then, a dynamical exponent adjustingalgorithm is developed, which takes the weights obtanied by quick dynamicalalgorithm as initial iteration value. The result demonstrate the advantage of theproposed algorithm.
出处
《南方冶金学院学报》
1997年第4期327-333,共7页
Journal of Southern Institute of Metallurgy
关键词
神经网络
模糊联想记忆
学习算法
吸引域
Neural network, Fuzzy associative memory, Learning algorithm, Basinof attraction
