Wang Tao, He Zhijun, Yu Ruizhao. Symmetric Neural Network: Asymptotic Stability and Learning[J]. Acta Electronica Sinica, 1993, (8): 86-89.DOI:
对称神经元网络:渐近稳定性和学习
摘要
本文在分析对称神经元网络的渐近稳定性基础上
提出了一种非约束优化学习算法
保证训练样本成为稳定吸引子
具有一定大小的吸引域.理论上
我们证明了算法的收敛性以及形成的吸引域下界.计算机实验结果充分说明了学习算法的优越性.
Abstract
In this paper
we present an unconstrained optimal learning rule by analyzing asymptotic stability of the symmetric neural network. It ensures to store each training pattern with a nontrivial basin of attraction. Theoretically
we study convergence theorem of the learning algorithm and a lower bound of attraction basin. Computer simulations are reported to show its advantages.