Finite-time synchronization of memristor-based neural networks
Haibo Bao
1,2
,JuH.Park
∗
2
1. School of Mathematics, Southwest University, Chongqing 400715, China
E-mail: hbbao07@gmail.com
2. Nonlinear Dynamics Group, Department of Electrical Engineering, Yeungnam University, 280 Daehak-Ro, Kyongsan 712-749,
Republic of Korea
E-mail: jessie@ynu.ac.kr
Abstract: This paper investigates finite-time synchronization of memristor-based neural networks. The purpose of
the addressed problem is to design a suitable controller to realize the finite-time synchronization between the drive and
response systems. Unlike the previous works, such finite-time synchronization will be realized for memristor-based
neural networks which are with discontinuous right-hand side. Based on finite-time stability theory and nonsmooth
analysis in mathematics, sufficient conditions are derived to ensure the finite-time synchronization of memristor-based
neural networks. An example is used to demonstrate the correctness of the main results.
Key Words: Finite-time synchronization, drive-response systems, memristor, Filippov’s solution, nonlinear control
1 INTRODUCTION
The concept of memristor (a contraction of memory and re-
sistor ) was first introduced by Chua in 1971 [1]. With the
invention of the first practical memristor device about 40
years later, it has been proved to be the fourth ideal electri-
cal circuit element [2, 3].
Since the pioneering work of Pecora and Carroll about syn-
chronization control of chaotic systems, neural networks
have been investigated due to its potential application in
secure communication, image processing, pattern recogni-
tion, associative memory.
In recent years, the dynamical analysis and synchroniza-
tion analysis of memristor-based neural networks have be-
come hot topics [4–8]. For instance, the Mitaag-Leffler
synchronization of memristor-based fractional-order neu-
ral networks was studied by Chen, Zeng and Jiang in
[6]. The problem of exponential synchronization and anti-
synchronization of memristor-based neural networks with
time-varying delays was disscussed in [8].
However, most of the studies relative to synchronization
of memristor-based neural networks are an infinite-time
asymptotical process, that is, only when the time tends to
infinity, the drive–response systems can reach synchroniza-
tion. And in theory, this will not occur in a finite time. So
we should further study the memristor-based neural net-
works, i.e., conduct finite-time synchronization.
Finite-time synchronization means optimal. It can im-
pose efficiency greatly in secure communication. By us-
∗
Corresponding author. This work was jointly supported by
the National Natural Science Foundation of China under Grant No.
61203096, the Chinese Postdoctoral Science Foundation under Grant
2013M513924, the Fundamental Research Funds for Central Univer-
sities XDJK2013C001 and the scientific research support project for
teachers with doctor’s degree, Southwest University under Grant No.
SWU112024. This work was also supported by 2014 Yeungnam Uni-
versity Research Grant (Chunma Chair Professorship).
ing finite-time synchronization technique, we can recover
the transmitted signals in a setting time, while other syn-
chronization techniques require an infinite time. Up to
now, there are some papers about finite-time synchroniza-
tion of neural networks and complex networks [9, 10]. To
the best of the authors’ knowledge, there are few papers
dealing with the finite-time synchronization of memristor-
based neural networks. Based on the above discussion,
the main objective of this paper is to address the finite-
time of memristor-based neural networks. By means of
the finite-time stability, nonlinear control theory and nons-
mooth analysis method, finite-time synchronization condi-
tions are given.
This paper is organized as follows. In Section 2, we formu-
late the problems and give some preliminaries. In Section
3, sufficient criteria are established to force the response
system to synchronize with the drive system. A numerical
example is given to prove the correctness of the main re-
sults in Section 4. Finally, some conclusions are made in
Section 5.
2 NOTATIONS and PRELIMINARIES
In this section, some elementary notations and lemmas are
introduced which play an important role in the proof of the
main results in Section 3.
Notation: Throughout this paper, ℝ
𝑛
denotes the 𝑛-
dimensional Euclidean space. The superscript “𝑇 ” denotes
vector transposition. ∣∣⋅∣∣is the Euclidean norm in ℝ
𝑛
.
Consider the following memristor-based neural networks
as the drive system:
˙𝑥
𝑖
(𝑡)=−𝑐
𝑖
𝑥
𝑖
(𝑡)+
𝑛
∑
𝑗=1
𝑎
𝑖𝑗
(𝑥
𝑗
(𝑡))𝑓
𝑗
(𝑥
𝑗
(𝑡)) + 𝐼
𝑖
, (1)
where 𝑖 =1, 2, ⋅⋅⋅ ,𝑛,𝑡 ≥ 0,𝑛is the number of units in a
neural network, 𝑥(𝑡)=(𝑥
1
(𝑡), ⋅⋅⋅ ,𝑥
𝑛
(𝑡))
𝑇
,𝑥
𝑖
(𝑡) denotes
1732
978-1-4799-7016-2/15/$31.00
c
2015 IEEE