时变时滞细胞神经网络的周期运动的稳定性  

Stability of Recurrent Cellular Neural Networks with Variable Time Delays

作　　者：郑伟范[1] 龙兰[2] 张继业[1] 张卫华[1] 

机构地区：[1]西南交通大学牵引动力国家实验室,四川成都610031 [2]重庆科技学院数理系,重庆400042

出　　处：《西南交通大学学报》2008年第2期222-226,共5页Journal of Southwest Jiaotong University

基　　金：国家自然基金资助项目(1077215250525518);教育部留学回国人员科研启动基金

摘　　要：利用M-矩阵理论和矢量Lyapunov函数方法,研究变时滞周期运动细胞神经网络的全局指数稳定性.在放松该类神经网络激活函数的有界性、单调递增性、可微性及Lipschitz连续等条件下,得到了该类神经网络周期解的存在性与全局指数稳定的代数判据.该判据基于神经网络激活函数满足的条件,利用连接权值矩阵及阻尼系数矩阵构造测试矩阵,根据测试矩阵是否为M-矩阵判定系统周期解的存在性与全局指数稳性.The global exponential stability of a class of recurrent cellular neural networks with variable time delays was studied using M-matrix theory and vector Lyapunov methods. Without assuming the boundedness, monotonicity, differentiability and Lipschitz continuity of the active functions, an algebraic criterion to ensure existence and globally m exponential stability of periodic solutions was obtained. From the weight matrix and damping coefficient matrix of the neural networks, a test matrix was constructed based on the conditions satisfied by the active functions. A recurrent cellular neural network has periodic and globally exponential stable solutions if the test matrix is an M-matrix.

关 键 词：神经网络 时变时滞 周期解 指数稳定性 M-矩阵 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]