摘要
本文利用神经元交互作用函数描述拓扑特征映射神经网络，探讨了这种网络的学习收敛性．本文首先给出一个网络收敛的一般性结论，并利用该结论证明网络输入满足平均分布时的收敛性．由此可进一步得到Ｋｏｈｏｎｅｎ网络自组织学习的收敛性．本文的结果修正并拓广了关于自组织学习收敛性已有的一些结果。
By defining the parameters of the neuron neighborhood interactions, the self organizing learning algorithm is extended to the more general case. Then a theorem on the topology preserving neural network's convergency is presented, by which a rigorous proof of the convergency of one dimensional neural networks with uniformly distributed input is presented. This paper revises and extends the existing results on the self organizing learning, and provides a new method for further proving the convergency of topology preserving neural networks completely.
出处
《计算机研究与发展》
EI
CSCD
北大核心
1997年第2期99-106,共8页
Journal of Computer Research and Development
关键词
神经网络
突触权值
自组织学习
收敛性
neural networks, synaptic weights, self organizing learning, convergency
