Impulsive effects on stability of discrete-time complex-valued neural
network s with both discrete and distributed time-varying delays
Qiankun Song
a,
n
, Zhenjiang Zhao
b
, Yurong Liu
c,d
a
Department of Mathematics, Chongqing Jiaotong University, Chongqing 400074, China
b
Department of Mathematics, Huzhou Teachers College, Huzhou 313000, China
c
Department of Mathematics, Yangzhou University, Yangzhou 225002, China
d
Communication Systems and Networks (CSN) Research Group, Faculty of Engineering, King Abdulaziz University, Jeddah 21589, Saudi Arabia
article info
Article history:
Received 26 March 2015
Received in revised form
26 April 2015
Accepted 2 May 2015
Communicated by Z. Wang
Available online 14 May 2015
Keywords:
Complex-valued neural networks
Discrete-time
Discrete time-varying delays
Distributed time-varying delays
Stability
Impulsive effects
abstract
In this paper, the impulsive effects on the stability of discrete-time complex-valued neural networks
with both discrete and distributed time-varying delays are investigated. Several sufficient conditions are
obtained ensuring the existence, uniqueness and global exponential stability of equilibrium point for the
considered neural networks by the idea of vector Lyapunov function method, M-matrix theory and
analytic technique. Moreover, the estimation for exponential convergence rate index is proposed. An
example with simulations is provided to verify the effectiveness of the obtained results.
& 2015 Elsevier B.V. All rights reserved.
1. Introduction
As is well known, time delays are often encountered unavoid-
ably in many practical systems such as automatic control systems,
population models, inferred grinding models, neural networks,
and so on [1,2]. They describe a kind of ubiquitous phenomenon
present in real systems where the rate of change of the state
depends on not only the current state of the system but also its
state at some time in history. Moreover, the existence of time
delays may lead to the instability or bad performance of systems
[3,4]. So, it is of prime importance to consider the delay effects on
the dynamical behavior of systems [5,6]. Over the past few
decades, neural networks with various types of delays have been
widely investigated, and many important and interesting results
on stability have been reported for delayed neural networks, for
example, see [7–17] and references therein.
On the other hand, an impulsive phenomenon exists uni versall y
in a wide variety of evolutionary processes where the state is cha-
nged abruptly at certain moments of time, involvin g such fields as
chemical tec hnology, population dynamics, physics and economics
[18]. It has also been shown that an impulsive phenomenon exists
likewise in neural networks [19]. For instance, during the imple-
mentation of electronic networks, when a stimulus from the body or
the external environ ment is received by receptors, the electrical
impulses will be convey ed to the neural net and an impulsive
phenomenon which is called impulsive perturbations arises naturally
[20]. The impulsive perturbation of neural networks can affect the
dynamical behaviors of the neural networ ks, same as time delays
effect [21] . There fore, it is necessary to consider both impulsive effect
and delay effect on dynamical behaviors of neural networks. There
are man y results on delayed neur al netw orks with impulsive
perturbations, for example, see [1 9–22] and references therein.
In the past few years, complex-valued neural networks (CVNN)
hav e been e xtensiv ely inv estigated o wing to man y successful appli-
cations of recurrent neural networks are involved in the sense of
complex-valued inputs. In fact, CVNN make it possible to solve some
problems which cannot be solved with their real-valued counter-
parts. For example, the exclusive OR (XOR) problem and the detec-
tion of symmetry problem cannot be solv ed with a single real-v alued
neuron, but they can be solved with a single complex-valued neuron
with the orthogonal decision boundaries, which reveals the potent
computational power of comple x-v a lued neurons [23].Hence,itis
very important to study about the dynamical properties of CVNN. As
pointed out in [24], CVNN is different and has more complicated
dynamical properties, compared with real-valued neural networks.
In real-valued neural networks, the activation functions are preferred
to be smooth and bounded. However, analytic and bounded
Contents lists available at ScienceDirect
journal homepage: www.els evier.com/locate/n eucom
Neurocomputing
http://dx.doi.org/10.1016/j.neucom.2015.05.020
0925-2312/& 2015 Elsevier B.V. All rights reserved.
n
Corresponding author.
E-mail addresses: qiankunsong@163.com (Q. Song),
zhaozjcn@163.com (Z. Zhao), liuyurong@gmail.com (Y. Liu).
Neurocomputing 168 (2015) 1044 – 1050