Elsevier

Neurocomputing

Volume 106, 15 April 2013, Pages 77-85
Neurocomputing

Exponential synchronization of coupled fuzzy neural networks with disturbances and mixed time-delays

https://doi.org/10.1016/j.neucom.2012.10.018Get rights and content

Abstract

This paper focuses on the exponential synchronization problem of coupled fuzzy neural networks with disturbances and mixed time-delays. For the network under study, the effects of both random and vague factors are considered. By stochastic analysis techniques, we establish sufficient conditions for the coupled fuzzy neural networks to be exponentially synchronized in the mean square. It is demonstrated that the network synchronizability is largely dependent on the coupling structure of such network. Moreover, the information exchange network needs not to be undirected or strongly connected. Finally, numerical simulations are given to verify the usefulness and effectiveness of our results.

Introduction

Recently, neural networks have been extensively applied in many areas, such as pattern recognition, image processing, secure communication, automatic control and associative memory, etc. [1], [2], [3], [4], [5], [6]. These applications greatly depend on the dynamical behaviors of networks. It is worth noting that realistic systems are often affected by some stochastic disturbances which may give rise to instability and poor performance. Considerable efforts [7], [8], [9], [10], [12], [13], [14], [15], [20], [24], [25], [27], [36] have been made to explore the mechanism of random noises.

In addition, it was pointed out in [5] that the combination of a set of neural networks could achieve higher level information processing. It is also an important issue to understand how the collective behaviors emerge from coupled systems [10]. As a typical collective behaviors, synchronization has gained increasing research interests [9], [10], [11], [12], [15], [16], [17], [18], [19], [23], [24], [25], [26], [28], [29], [31], [32], [33]. Taking constant, discrete-delay and distributed-delay couplings into consideration, the authors of [11] derived the global synchronization criteria for coupled neural networks. In [12], this result was extended to a discrete-time model with stochastic nonlinearities and mixed time-delays.

Generally, the exact modeling of practical systems is not always available. In traditional neural networks, present knowledge cannot be fully utilized to improve the network structure and speed up knowledge acquiring [20]. In this case, pre-assuming a certain model for dynamical system may lead to unpractical results. Nowadays, it has been wildly recognized that fuzzy logic theory is an efficient approach to deal with such problems. Since fuzzy logics were translated into cellular neural network [21], fuzzy cellular neural network (FCNN) has received widespread attentions [5], [20], [21], [22], [23], [24], [25], [26], [27]. Furthermore, it was proved in [25] that fuzziness could bring about more complexities. Therefore, studying the dynamical behaviors of fuzzy systems has both theoretical and practical significance.

However, the fuzzy logics and network synchronizability defined as the ability that system structure guarantees synchronization [19] are rarely considered in coupled stochastic neural networks with mixed time-delays. All of these motivate our work. In this paper, we study the exponential synchronization problem of coupled fuzzy neural networks (CFNNs) with disturbances and mixed time-delays. Time-delays are mixed in the sense that both discrete delay and distributed delay are involved. The former is employed to describe the propagation delay among neurons, and the latter is utilized to characterize the spatial nature of neural networks [9]. We aim to develop some sufficient criteria that guarantee the mean square exponential synchronization of CFNNs, and study the effects of network structure on network synchronizability.

This paper is organized as follows. In Section 2, the model is introduced. Sufficient conditions that ensure the mean square exponential synchronization of CFNNs are stated in Section 3. In Section 4, two numerical examples are given, and conclusion is made in Section 5.

Notations: Rn and Rm×n denote the n-dimensional Euclidean space and the set of m×n real matrices, respectively. For vector xRn,x=xTx,|x|=(|x1|,,|xn|)T. For real symmetric matrix ARn×n,A<0 (A>0) means A is negative definite (positive definite). We let C([τ,0],Rn) denote the family of continuous functions φ from [τ,0] to Rn with norm φ=supτξ0|φ(ξ)|. Let (Ω,F,P) be a complete probability space. Denote by LF02([τ,0],Rn) the family of all F0-measurable C([-τ,0],Rn)-valued random variables satisfying supτξ0E{φ(ξ)2}< where E{·} stands for the mathematical expectation operator. λmax(A) and λmin(A) are separately the maximal and minimum eigenvalue of square matrix A. det(A) denotes the determinant of A. diag(a1,,an) is a diagonal matrix with diagonal elements a1,,an. Denote by In the n×n identity matrix. represents the Kronecker product of matrices.

Section snippets

Problem formulations and preliminaries

The fuzzy cellular neural network with Markovian switching parameters is given bydym(t)=c˜m(γ(t))ym(t)+l=1na˜ml(γ(t))f˜l(yl(t))+l=1nb˜ml(γ(t))g˜l(yl(tτ(γ(t)))+l=1nd˜ml(γ(t))tτ(γ(t))th˜l(yl(ξ))dξ+l=1nu˜ml(γ(t))vl(t)+l=1nϕmlg˜l(yl(tτ(γ(t))))+l=1nψmlg˜l(yl(tτ(γ(t))))+l=1nδmlvl(t)+l=1nθmltτ(γ(t))th˜l(yl(ξ))dξ+l=1nπmlvl(t)+l=1nλmltτ(γ(t))th˜l(yl(ξ))dξdt,m=1,,n,where y(t)=(y1(t),,yn(t))TRn and v(t)=(v1(t),,vn(t))TRn are, respectively, the state variables and external inputs

Main results

In this section, some criteria are established in terms of linear matrix inequalities (LMIs) that guarantee the mean square exponential synchronization of CFNNs.

In order to prove that the CFNNs (2) (or (3)) is exponentially synchronized in the mean square, we need to ensure the exponential mean square stability of (4).

For simplifying representation, we denote γ=maxkS(|γkk|), τ^=maxkS(τk), τˇ=minkS(τk), F˜=diag(max(|f1|,|f1+|),,max(|fn|,|fn+|)), F=IN1F˜, G˜=diag(max(|g1|,|g1+|),,max(|g

Numerical examples

In this section, two numerical examples are given to verify our results obtained in Section 3.

Consider the CFNNs (2) (or (3)) with Markovian switching between two modes (k=1,2), in which xi(t)=(xi1(t),xi2(t))T, f˜(xi(t))=g˜(xi(t))=h˜(xi(t))=((|xi1(t)+1||xi1(t)1|)/2,(|xi2(t)+1||xi2(t)1|)/2)T, σ˜(t,xi(t),xi(tτk),k)=R˜kxi(t)+S˜kxi(tτk), v(t)=(sin(t),cos(t))T,A˜1=0.30.20.10.5,A˜2=0.50.10.20.3,B˜1=0.20.30.40.3,B˜2=0.30.10.10.2,D˜1=0.10.20.30.2,D˜2=0.20.10.10.1,C˜1=diag(0.2,0.1),C˜2=

Conclusion

In this paper, the network synchronizability of CFNNs with stochastic noises and mixed time-delays is discussed under a mild topology-based assumption. It is showed that coupling strength and spanning tree play a key role in synchronization process. LMI-based conditions are established to guarantee the exponential synchronization in the mean square sense for CFNNs.

Based on the properties of energy, passivity theory was proposed to analyze the input and output in neural networks [36]. We will

Guan Wang received his M.S. degree in Systems Engineering from Huazhong University of Science and Technology, Wuhan, China, in 2009. He is currently working toward his Ph.D. degree at the Department of Control Science and Engineering, Huazhong University of Science and Technology. His current research interests include neural networks and complex systems.

References (36)

  • A. Langville et al.

    The Kronecker product and stochastic automata networks

    J. Comput. Appl. Math.

    (2004)
  • S. Zhu et al.

    Passivity analysis of stochastic delayed neural networks with Markovian switching

    Neurocomputing

    (2011)
  • L. Chua et al.

    Cellular neural networkstheory

    IEEE Trans. Circuits Syst.

    (1988)
  • H. Jiang et al.

    Boundedness and stability for nonautonomous bidirectional associative neural networks with delay

    IEEE Trans. Circuits Syst. Express Briefs

    (2004)
  • C. Chang et al.

    A recurrent fuzzy coupled cellular neural network system with automatic structure and template learning

    IEEE Trans. Circuits Syst. Express Briefs

    (2006)
  • Y. Shen et al.

    Noise-induced stabilization of the recurrent neural networks with mixed time-varying delays and Markovian-switching parameters

    IEEE Trans. Neural Networks

    (2007)
  • Y. Shen et al.

    Almost sure exponential stability of recurrent neural networks with markovian switching

    IEEE Trans. Neural Networks

    (2009)
  • J. Lu et al.

    Pinning stabilization of linearly coupled stochastic neural networks via minimum number of controllers

    IEEE Trans. Neural Networks

    (2009)
  • Cited by (0)

    Guan Wang received his M.S. degree in Systems Engineering from Huazhong University of Science and Technology, Wuhan, China, in 2009. He is currently working toward his Ph.D. degree at the Department of Control Science and Engineering, Huazhong University of Science and Technology. His current research interests include neural networks and complex systems.

    Quan Yin received the B.S., M.S. degrees in Industrial Electrical Automation, and Ph.D. degree in Control Theory and Control Engineering from Huazhong University of Science and Technology, Wuhan, China, in 1990, 1995 and 2001, respectively. He was a Post-Doctoral Fellow with the Huazhong University of Science and Technology from July 2001 to June 2003. He is currently a Associate Professor with the Department of Control Science and Engineering, Huazhong University of Science and Technology, Wuhan, China. He is the author of more than 30 research papers. His current research interests include intelligent control and embedded control system.

    Yi Shen received the Ph.D. degree from Huazhong University of Science and Technology, Wuhan, China, in 1998. He was a Postdoctoral Research Fellow at the Huazhong University of Science and Technology during 1999–2001. He joined the Department of Control Science and Engineering, Huazhong University of Science and Technology, and became an Associated Professor in 2001 and was promoted to Professor in 2005. From 2006 to 2007 he was a Postdoctoral Research Fellow at the Department of Mechanical and Automation Engineering, the Chinese University of Hong Kong, Hong Kong. He has authored over 80 research papers. His main research interests lie in the fields of stochastic control and neural networks.

    The project was supported by the State Key Program of National Natural Science of China under Grant no. 61134012.

    View full text