非线性连续神经网络全局指数稳定的定量特征

THE QUANTITATIVE PROPERTIES OF EXPONENTIAL STABILITY OF NONLINEAR CONTINUOUS NEURAL NETWORK

引入一个特征函数用以定量刻画最优化计算神经网络的全局指数稳定性。利用该特征函数得到了一族范数下神经网络全局指数稳定的判定条件及下三角神经网络全局指数稳定的充分必要条件。所得结果推广了Bouzerdoum,Killy,Sugawara等人的结论。

In the paper, a characteristic function is introduced and used to characterize quan-titatively global exponential stability of nonlinear continuous neural network . Some sufficient conditions for the network to be exponentially stable under all monotone norms are presented, and the sufficient and necessary condition for lower-triangle neural network to be globally ex-ponentially stable is obtained also.