Message Board

Dear readers, authors and reviewers,you can add a message on this page. We will reply to you as soon as possible!

2017 Volume 39 Issue 6
Article Contents

Tian-wei MU, Ruo-feng RAO. An Application of the Fixed Point Theory in Stability Analysis of Delayed Bi-Directional Associative Memory (BAM) Neural Networks[J]. Journal of Southwest University Natural Science Edition, 2017, 39(6): 5-9. doi: 10.13718/j.cnki.xdzk.2017.06.002
Citation: Tian-wei MU, Ruo-feng RAO. An Application of the Fixed Point Theory in Stability Analysis of Delayed Bi-Directional Associative Memory (BAM) Neural Networks[J]. Journal of Southwest University Natural Science Edition, 2017, 39(6): 5-9. doi: 10.13718/j.cnki.xdzk.2017.06.002

An Application of the Fixed Point Theory in Stability Analysis of Delayed Bi-Directional Associative Memory (BAM) Neural Networks

More Information
  • Corresponding author: Ruo-feng RAO
  • Received Date: 28/07/2016
    Available Online: 20/06/2017
  • MSC: O177.91, O193

  • In this paper, by formulating a contraction mapping in the product space, the authors derive a global exponential stability criterion for BAM neural networks with time-delays. Different from the existing literature, the newly-obtained criterion can be easily verified by the computer matlab LMI toolbox. Moreover, a numerical example is presented to illustrate the effectiveness of the proposed method.
  • 加载中
  • [1] KOSKO B. Bidirectional Associative Memories[J]. IEEE Trans Syst Man Cybern, 1988, 18(1): 49-60. doi: 10.1109/21.87054

    CrossRef Google Scholar

    [2] RAO R F, ZHONG S M, WANG X R. Stochastic Stability Criteria with LMI Conditions for Markovian Jumping Impulsive BAM Neural Networks with Mode-Dependent Time-Varying Delays and Nonlinear Reaction-Diffusion[J]. Communications in Nonlinear Science and Numerical Simulation, 2014, 19(1): 258-273. doi: 10.1016/j.cnsns.2013.05.024

    CrossRef Google Scholar

    [3] RAO R F, WANG X R, ZHONG S M, et al. LMI Approach to Exponential Stability and Almost Sure Exponential Stability for Stochastic Fuzzy Markovian-Jumping Cohen-Grossberg Neural Networks with Nonlinear p-Laplace Diffusion[J]. Journal of Applied Mathematics, 2013, 2013: 1-21.

    Google Scholar

    [4] RAO R F, ZHONG S M, PU Z L. LMI-Based Robust Exponential Stability Criterion of Impulsive Integro-Differential Equations with Uncertain Parameters Via Contraction Mapping Theory[J]. Advances in Difference Equations, 2017, 2017: 1-16. doi: 10.1186/s13662-016-1057-2

    CrossRef Google Scholar

    [5] 俸卫. T-S模糊马尔可夫跳跃时滞Cohen-Grossberg神经网络的几乎必然指数稳定性[J].西南师范大学学报(自然科学版), 2015, 40(5): 10-12.

    Google Scholar

    [6] 黄家琳, 饶若峰. Cohen-Grossberg神经网络的全局指数稳定性[J].西南大学学报(自然科学版), 2016, 38(2): 78-82.

    Google Scholar

    [7] RAO R F, PU Z L. LMI-Based Stability Criterion of Impulsive T-S Fuzzy Dynamic Equations Via Fixed Point Theory[J]. Abstract and Applied Analysis, 2013, 2013: 1-9.

    Google Scholar

  • 加载中
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Article Metrics

Article views(2046) PDF downloads(307) Cited by(0)

Access History

Other Articles By Authors

An Application of the Fixed Point Theory in Stability Analysis of Delayed Bi-Directional Associative Memory (BAM) Neural Networks

    Corresponding author: Ruo-feng RAO

Abstract: In this paper, by formulating a contraction mapping in the product space, the authors derive a global exponential stability criterion for BAM neural networks with time-delays. Different from the existing literature, the newly-obtained criterion can be easily verified by the computer matlab LMI toolbox. Moreover, a numerical example is presented to illustrate the effectiveness of the proposed method.

  • 1988年,文献[1]首次引进双向联想记忆神经网络(BAM神经网络). BAM神经网络能广泛应用于人工智能、图像恢复、信号与图像处理、组合优化、联想记忆等诸方面,这些成功的应用很大程度上依赖于系统本身是否具有某种稳定性.本文考虑一类时滞BAM神经网络的全局指数型稳定性判据.一直以来,人们通常考虑用李雅普诺夫方法及其它方法解决时滞神经网络或动力系统的稳定性判别问题[1-5],但每一种方法都有其局限性,李雅普诺夫方法也不例外.于是人们有时也考虑用其它方法来弥补李雅普诺夫方法的不足,其中不动点方法是人们考虑的备选方法.最近,不动点方法应用到了神经网络的稳定性分析,获得了一系列新的稳定性判定准则[6-7].而我们发现,不动点方法较少应用到BAM神经网络稳定性分析中.事实上,不动点方法要运用到BAM神经网络稳定性分析中的确存在一些数学上的困难.本文拟构造一个乘积空间上的压缩映射来克服数学上的困难,将运用压缩映像原理来获得其稳定性判据.由于选择方法不同,结论不同于以往结果.

    考虑以下时滞BAM神经网络:

    其中:

    激活函数:

    时滞τ(t),h(t)满足0≤τ(t),h(t)≤τ. AB皆为正定对角矩阵,为神经元势能恢复参数矩阵.以下全文假设:

    以及:

    (A1) 存在正定对角矩阵F,使得

    (A2) 存在正定对角矩阵G,使得

    为方便起见,对于向量u=(u1u2,…,un)T$\mathbb{R}$ n及矩阵M=(mij)n×n,我们记:

    对向量uv$\mathbb{R}$n,向量不等式uv等价于其分量不等式uivi (∀i=1,2,…,n).

    本文的主要结果是:

    定理1  假设存在常数0<λ<1,以下LMI条件成立:

    则系统(1) 是全局均方指数型稳定的.

      首先定义乘积空间Ω=Ω1×Ω2如下:设函数空间Ωi(i=1,2) 由满足下列3个条件的函数qi(t):[-τ,∞) $\to \mathbb{R}$ Rn构成:

    (a) qi(t)在t∈[0,+∞)上连续;

    (b) q1(t)=ξ(t),q2(t)=η(t) (∀t∈[-τ,0]);

    (c)当t→∞时,eγtqi(t)→0∈$\mathbb{R}$ n,其中γ>0是正的常数,满足γ<min{λminAλminB}.

    再定义空间Ω上的距离如下:

    其中:

    这里${{{\mathit{\boldsymbol{\bar{q}}}}}_{i}}$Ωi${{{\mathit{\boldsymbol{\tilde{q}}}}}_{i}}$Ωi (i=1,2).则不难证明Ω是上述距离下的完备度量空间.

    下面分3步来完成定理1的全部证明:

    步骤1  构造Ω上的映射.

    为系统(1) 的解.则对任意t≥0,我们有

    等式两边积分,则有

    类似于(4) 式,我们有

    于是构造Ω上的映射P如下:

    步骤2  证明映射P是自射的,即对任意$\left( \begin{align} & \mathit{\boldsymbol{x}}\left( t \right) \\ & \mathit{\boldsymbol{y}}\left( t \right) \\ \end{align} \right)$$p\left( \begin{align} & \mathit{\boldsymbol{x}}\left( t \right) \\ & \mathit{\boldsymbol{y}}\left( t \right) \\ \end{align} \right)\in \mathit{\Omega} $换而言之,证明$p\left( \begin{align} & \mathit{\boldsymbol{x}}\left( t \right) \\ & \mathit{\boldsymbol{y}}\left( t \right) \\ \end{align} \right)$满足条件(a)-(c).

    事实上,由(5),(6) 式,显然条件(a)和(b)可以满足.

    而满足条件(c)只要下式成立:

    t→∞时,显然:

    下面证明

    由eγtx(t)→0知,对任意给定的ε>0,存在相应的正数t*τ,使得

    其中μ=(1,1,…,1)T${{\mathbb{R}}^{n}}$.由条件(A1) 和(A2),有

    一方面,我们有

    另一方面,显然存在正数a0,使得|C|a0I,这里I表示单位矩阵.从而

    则由(9)-(11) 式就导出了(8) 式.类似可证得

    显然由(8) 和(12) 式知(7) 式成立,从而条件(c)满足.因此我们证得:对任意给定的$\left( \begin{align} & \mathit{\boldsymbol{x}}\left( t \right) \\ & \mathit{\boldsymbol{y}}\left( t \right) \\ \end{align} \right)\in \mathit{\Omega} $,有$p\left( \begin{align} & \mathit{\boldsymbol{x}}\left( t \right) \\ & \mathit{\boldsymbol{y}}\left( t \right) \\ \end{align} \right)\in \mathit{\Omega} $.

    步骤3  证明PΩ上的压缩映射.

    事实上,对任给$\left( \begin{array}{l} \mathit{\boldsymbol{x}}\left( t \right)\\ \mathit{\boldsymbol{y}}\left( t \right) \end{array} \right), \left( \begin{array}{l} \mathit{\boldsymbol{\bar x}}\left( t \right)\\ \mathit{\boldsymbol{\bar y}}\left( t \right) \end{array} \right) \in \Omega $, 有

    从而

    其中A-1B-1分别是AB的逆矩阵.所以PΩ $ \to $ Ω是压缩映射.因此存在PΩ上的不动点$\left( \begin{array}{l} \mathit{\boldsymbol{x}}\left( t \right)\\ \mathit{\boldsymbol{y}}\left( t \right) \end{array} \right)$, 同时它也是系统(1) 的解,当t→∞时,满足${{\rm{e}}^{rt}}\left( \begin{align} & \mathit{\boldsymbol{x}}\left( t \right) \\ & \mathit{\boldsymbol{y}}\left( t \right) \\ \end{align} \right)\to 0\in {{\mathbb{R}}^{2n}}$, 这就证明了定理1.

    例1  给系统(1) 配置参数如下:令n=2,时滞上限τ=8.5,以及:

    则套用计算机Matlab LMI工具箱可获得可行性结果[3]

    显然0<λ<1,从而由定理1知系统(1) 是全局均方指数型稳定的.

    本文通过构造乘积空间上的压缩映像,获得了一类时滞BAM神经网络的全局指数型稳定判据.特别地,乘积空间上的压缩映像的构造法是以前相关文献所没有的设想和方法,同时构造合适的乘积空间上的距离也是证明定理的关键技巧.总之,通过构造乘积空间上的压缩映像,我们克服了时滞BAM神经网络数学模型在数学上的困难,最终获得了稳定性判据.本文也为进一步的研究提供了方法上的一些启迪[4],数值实例证实了所述方法的有效性.

Reference (7)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return