摘要
提出两种功能互相不同的神经细胞组成的复合神经元网络(CNN)模型;导出一种特殊结构的CNN的并行动力学;而且证明了它的稳定性。在这些结果基础上,得到快速的假逆矩阵学习算法。计算机仿真试验证实学习算法与动力学稳定性的正确性,并表现出良好的容错性能与存储容量。
A complex neural network (CNN) , which consists of two types of nerve cell, differing functionally from each other, is suggested. Parallel dynamics for a special architecture of CNN is derived and its stability is proved. Based on the above-mentioned result, a fast pseudo-inverse matrix learning algorithm is obtained for the CNN. The validity of the learning algorithm and the dynamical stability are confirmed by the computer simulated experiments.
出处
《生物物理学报》
CAS
CSCD
北大核心
1992年第1期165-173,共9页
Acta Biophysica Sinica
关键词
神经元网络
并行动力学
学习算法
Neural network, Parallel dynamics, learning algorithm.