随机Hopfield神经网络渐进行为的变互式凸组合方法  

Reciprocally Convex Approach to Asymptotic Behavior of Stochastic Hopfield Neural Networks

作　　者：陈无及 李青松[1] 

机构地区：[1]长沙理工大学数学与计算科学学院,长沙

出　　处：《应用数学进展》2014年第2期70-77,共8页Advances in Applied Mathematics

摘　　要：本文通过构造合适Lyapunov-Krasovskii泛函,采用变互式凸组合方法、自由权矩阵方法、随机分析技巧、线性矩阵不等式(LMI)方法,研究了一类随机Hopfield神经网络的随机一致最终有界性、随机吸引子的存在性以及均方指数的稳定性。我们得到了随机Hopfield神经网络的新的结果和时滞依靠的充分条件。最后,应用MATLAB数值模拟检验来得到了理论的有效性。In this paper, theproblems of stochastic uniformly ultimate boundedness, the existence of stochasticattractor and mean square exponential stability for stochastic Hopfield neuralnetworks are studied by constructing proper Lyapunov-Krasovskii functional,utilizing the reciprocally con- vex approach, the free-weighting matrix method,some stochastic analysis techniques and linear matrix inequalities technique (LMI). We deduce novel results anddelay-dependent sufficient conditions for stochastic Hopfield neuralnetworks. Finally, numerical simulations are given on MATLAB to verify theeffectiveness of the gained results.

关 键 词：HOPFIELD神经网络 有界性 均方指数稳定 变互式凸组合方法 

分 类 号：TP1[自动化与计算机技术—控制理论与控制工程]