Brief Papers
Finite-time stability analysis of discrete-time fuzzy Hopfield neural
network
Jianjun Bai
a,
n
, Renquan Lu
a
, Anke Xue
a
, Qingshan She
a
, Zhonghua Shi
b
a
Institute of Information Science and Control, Hangzhou Dianzi University, Hangzhou 310018, PR China
b
China Petroleum Longhui Automation Engineering Co. Ltd, PR China
article info
Article history:
Received 17 November 2014
Received in revised form
13 January 2015
Accepted 29 January 2015
Communicated by Guang Wu Zheng
Available online 13 February 2015
Keywords:
Fuzzy Hop field neural network
Finite-time stability
Generalized activation function
abstract
The finite-time stability analysis of discrete-time fuzzy Hopfield neural networks is studied in this paper.
Firstly, the concept of finite-time stability is generalized to the fuzzy neural networks. And then by the
Lyapunov approach and linear matrix inequality technique, a sufficient condition for the system to be
finite-time stable is proposed, based on which, the finite-time stability condition for the system with
norm bounded uncertainties is also given. Finally a numerical example is given to illustrate the
effectiveness of the proposed approach.
& 2015 Elsevier B.V. All rights reserved.
1. Introduction
The well-known Hopfield neural network has been widely
investigated since it was proposed by Hopfield in [1]. It has been
applied successfully to many different fields such as combinatorial
optimization, signal processing and pattern recognition, see [2–10]
and the references therein. In practice, it is necessary to discretize
the continuous system to a discrete-time system when it is
implemented based on computer [11–16] and some results about
the discrete time Hopfield neural networks are available in the
literature [17–19].
On the other hand, it is impossible to avoid the vagueness in
the real world and the fuzzy theory is viewed as an appropriate
approach to deal with this problem [20,21]. So some researchers
began to combine the low level information processing capability
of neural networks with the high level information processing
capability of fuzzy systems to improve the performance of the
neural networks [22]. Due to the combination ways of fuzzy logic
and neural network, fuzzy neural networks are classified into two
types: neural networks incorporating some fuzzy logic or neural
networks based on fuzzy inference [23]. The latter one is more
attractive since it is based on the T–S fuzzy models, which blend
some linear models via nonlinear membership functions to repre-
sent a complex nonlinear system [24]. Many problems of the fuzzy
neural networks have also been studied [25,26].
The classical control theory focuses mainly on the asymptotic
behavior of the systems, which deals with the asymptotic property of
system trajectories over an infinit e time interval. However, as pointed
out in [27], in some case, it is more reasonable to make the system
tra ject ories not to ex ceed a certain threshold in a finite-time interval,
for example, large values of the state are not acceptable in the
presence of the saturati on, which is also called finite-time stability
(FTS) [28–31].Thefinite-time stability of the continuous system
is studied in [27] and the discrete-time case in [32]. The output
feedback control is considered in [33]. For the neural networks, the
finite time stability is firstly studied in [34],andthefinite-time
boundedness of delayed neural netw orks with Markovian jumping
parameters is investigated in [35,36].The
finit e-time estimation
problem is also studied in [37]. T o the best of our knowledge,
finite-time stability analysis of fuzzy Hopfield neural netw orks has
not been investigated in the literature, which motivates the study of
this paper .
The rest of this paper is organized as follows: Section 2
presents the problem formulation, some lemmas and definitions
are also given. The main results of this paper are given in Section 3.
An example is given in Section 4 to show the effectiveness of the
proposed approach and Section 5 is the conclusion.
Notation: Throughout this paper, N
0
denotes the set of natural
numbers, R
n
denotes the n dimensional real Euclidean space, C
denotes the complex plane, I
k
is the k k identity matrix, the
superscripts ‘T’ and ‘ 1’ stand for the matrix transpose and
inverse respectively, ‘
n
’ denotes the symmetric element in a
matrix.
λ
max
and
λ
min
denote the maximum and minimum eigen-
value of a matrix. W 4 0 ðW Z 0Þ means that W is real, symmetric
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/ locate/neucom
Neurocomputing
http://dx.doi.org/10.1016/j.neucom.2015.01.051
0925-2312/& 2015 Elsevier B.V. All rights reserved.
n
Corresponding author.
E-mail address: baijianjun@hdu.edu.cn (J. Bai).
Neurocomputing 159 (2015) 263–267