孙云霞
摘 要:基于离散时间状态观测,研究带Markov切换的随机Cohen-Grossberg神经网络稳定的问题.通过构造 Lyapunov函数,利用[Ito]微分公式、Borel-Cantellis引理及稳定性分析理论,得到非线性和线性系统几乎必然指数稳定的充分条件.最后,通过一个例子验证所得结果的可行性.
关键词:随机Cohen-Grossberg神经网络;Markov链;离散时间观测;Lyapunov函数;几乎必然指数稳定
中图分类号:O231 DOI:10.16375/j.cnki.cn45-1395/t.2020.04.012
0 引言
自Cohen[1]首次提出和研究Cohen-Grossberg神经网络模型(简称CGNNs)以来,神经网络在模式识别、并行处理、联想记忆、最优化计算等方面得到了较深入研究.而系统在其运行过程经常会出现随机突变现象,如外部环境的突然变化、大系统内部各子系统间连接方式的改变等都会引起系统结构的改变,而对于这种具有可变结构的系统常常利用Markov切换[2-3]系統模型来刻画.另外,噪声[4-5]能使一个不稳定的系统稳定,甚至使一个稳定的系统更稳定.由于在实际的系统中不可避免的存在噪声的因素,因此,在很多的文献中都考虑到它的影响.
众所周知,传统的反馈控制[σxt, rt, t]要求在所有的时间上对状态是进行连续观测的,在经济上是非常昂贵的,实际上连续情况的观测也可能无法实现.所以,毛学荣[6]提出了在离散时间观测的基础上设计一个反馈控制[σxδt, rδt, t],使其控制更加合理和实用,其中,[δt=tττ],[τ>0],[tτ]表示[tτ]的整数部分.通过选择正常数[τ],只需要观测[0],[τ],[2τ],[…]. 离散时间间隔[τ]的上界非常小,对于如何选择一段时间间隔,以使得反馈控制系统稳定,可以参考文献[6]. 然而,能否通过用[σxδt, rδt, t]替换,使带Markov切换的随机CGNNs具有几乎必然指数稳定性?这是本文研究的关键问题.
近年来,随机CGNNs的稳定性[7-9]研究吸引了大量学者,并取得了许多有意义的结果.目前关于带Markov切换的随机CGNNs的几乎必然指数稳定性[10]和基于离散时间状态观测的随机系统的几乎必然指数稳定性[6, 11-12]均已研究.但是,基于离散时间状态观测,研究带Markov切换的随机CGNNs的几乎必然(a.s)指数稳定性的却不多见.
本文在文献[6, 11-12]的启发下,研究如何通过加入一个反馈控制器使不稳定的随机CGNNs达到稳定性的问题.
1 准备知识
本文采用以下记号:记[Ω, F, Ftt≥0, P]为含有满足通常条件的代数流[Ftt≥0]的完备概率空间,令[Bt=B1t, B2t, …, BmtT]为定义于该空间上的[m]维标准布朗运动.
4 结论
本文利用[Ito]微分公式、Borel-Cantellis引理和稳定性分析理论对基于布朗运动和离散时间状态观测的带Markov切换的随机CGNNs的几乎必然指数稳定性进行了研究.最后得到判定其线性和非线性随机CGNNs几乎必然指数稳定的充分性条件.同时,对于间隔[τ],如果能指出一个较大的上界可以大大地降低控制成本.
参考文献
[1] COHEN M C. Absolute stability of global pattern formation and parallel memory storage by competitive neural networks[J]. IEEE Transactions on Systems,Man,and Cybernetics,1983,13:815-826.
[2] XING M L,DENG F Q. Adaptive cooperative tracking control of uncertain nonlinear multiagent systems with uncertain Markov switching communication graphs[J]. International Journal of Adaptive Control and Signal Processing,2019,33(10):1506-1523.
[3] CHAPPLE A G,PEAK T,HEMAL A. A novel Bayesian continuous piecewise liner log-hazard model,with estimation and inference via reversible jump Markov chain Monte Carlo[J]. Statistics in Medicine,2020,39(12):1766-1780.
[4] CHENG P,DENG F Q,YAO F Q. Almost sure exponential stability and stochastic stabilization of stochastic differential systems with impulses effects[J]. Nonlinear Analysis:Hybrid Systems,2018,30:106-117.
[5] 李创第,华逢忠,葛新广. Maxwell阻尼耗能多层结构在有界噪声激励下的随机解析分析[J]. 广西科技大学学报,2016,27(4):1-6,20.
[6] MAO X R. Almost sure exponential stabilization by discrete-time stochastic feedback control[J]. IEEE Transactions on Automatic Control,2016,61(6):1619-1624.
[7] CHENG P,SHANG L,YAO F Q. Exponential stability of Markovian jumping stochastic Cohen-Grossberg neural networks with mixed time delays and impulses[C]//Proceedings of the 34th Chinese Control Conference,2015,1793-1798.
[8] ZHU Q X,CAO J D. Robust exponential stability of Markovian jumping impulsive stochastic Cohen-Grossberg neural networks with mixed time delays[J]. IEEE Transactions on Neural Networks, 2010,21(8):1314-1324.
[9] RAKKIYAPPAN R,CHANDRASEKAR A,LAKSHMANAN S,et al. Exponential stability of Markovian jumping stochastic Cohen-Grossberg neural networks with mode-dependent probabilistic time-varying delays and impulses[J]. Neurocomputing,2014,131:265-277.
[10] SUN Y X,CHENG P,YAO F Q. Almost sure exponential stability of stochastic Cohen–Grossberg neural networks with Markovian jumping and impulses[C]//Proceedings of the 36th Chinese Control Conference,2017,1876-1881.
[11] DONG R. Almost sure exponential stabilization by stochastic feedback control based on discrete-time observations[J]. Stochastic Analysis Applications, 2018,36(4):561-583.
[12] 余佩琳,崔瑤,程培. 混杂随机系统基于离散时间状态观测的几乎必然指数稳定[J]. 广西科技大学学报,2019,30(2):115-120.
[13] MAO X R,YUAN C G. Stochastic differential equations with Markovian switching[M]. London:Imperial College Press,2006.
[14] SONG G F,ZHENG B C,LUO Q,et al. Stabilization of hybrid stochastic differential equations by feedback control based on discrete-time observations of state and mode[J]. IET Control Theory and Applications,2017,11(3):301-307.
(责任编辑:罗小芬、黎 娅)