Professional Documents
Culture Documents
com
Abstract
In this paper, several classes of impulsive Cohen–Grossberg neural networks with continuously distributed delays
are considered. Global exponential stability and robust global exponential stability of the equilibrium solution are
investigated by using Lyapunov function and integro-differential inequality. Moreover, sufficient conditions are also
given to guarantee the existence of --periodic solution and that all other solutions are convergent to it globally expo-
nentially. Finally, two examples are given to demonstrate the effectiveness of our results in this paper.
2007 Elsevier Ltd. All rights reserved.
1. Introduction
Cohen–Grossberg neural networks (CGNNs) were firstly introduced by Cohen–Grossberg [1] in 1983, which can be
described as follows:
" #
X
n
x_ i ðtÞ ¼ ai ðxi ðtÞÞ bi ðxi ðtÞÞ cij fj ðxj ðtÞÞ þ I i ; i ¼ 1; 2; . . . n;
j¼1
where n P 2 is the number of neurons in this networks, xi(t) is the state variable of the ith neuron, ai() > 0 represents an
amplification function, bi() is the appropriately behaved function, (cij)nn denotes the connection matrix in which cij
represents the connection strength from ith neuron to jth neuron, fj() is the non-linear input–output activation of
jth neuron, I = (I1, I2, . . . , In)T 2 Rn denotes input vector. Cohen–Grossberg neural networks have been used in many
fields such as pattern recognition, associative memory, signal processing and combinatorial optimization [2,3]. Such
applications heavily depend on the dynamic behaviors of Cohen–Grossberg neural networks. It is well known that
the stability and convergence are prerequisites to guarantee the performance of Cohen–Grossberg neural networks.
Therefore, it is of prime importance to analyze the dynamics of Cohen–Grossberg neural networks.
*
Corresponding author. Tel.: +86 21 34 202004; fax: +86 21 34 202030.
E-mail address: jglu@sjtu.edu.cn (J.G. Lu).
0960-0779/$ - see front matter 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.chaos.2007.11.022
Z.W. Ping, J.G. Lu / Chaos, Solitons and Fractals 41 (2009) 164–174 165
Due to the finite speeds of the switching of the amplifiers and transmission of signals in hardware implementation
[4], time delay should exist in electronic neural networks. On the other hand, it has also been shown that the process of
moving images requires the introduction of delay in signal transmitted through the networks [5]. However, Time delays
may lead to bifurcation, oscillation, divergence, or instability which may be harmful to a system. Thus, the study of
neural dynamics with consideration of delay problem becomes extremely important to manufacture high quality
Cohen–Grossberg neural networks. Recently, the stability of Cohen–Grossberg neural networks with discrete delays
has been studied, e.g. Refs. [6–23].
In addition, it is well known that in real world many evolutionary processes, particularly some biological systems such
as biological neural networks and bursting rhythm models in pathology are characterized by abrupt changes of states at
certain time. These changes are called to be impulsive phenomena, which are also included in many fields such as physics,
chemistry, population dynamics, optimal control, etc. However, impulsive effects may also lead to bifurcation, oscillation,
divergence, or instability which may be harmful to a system. Thus, the study of neural dynamics with consideration of
impulsive effects becomes extremely important to manufacture high quality Cohen–Grossberg neural networks. Recently,
the stability of impulsive Cohen–Grossberg neural networks with discrete delay has been studied in [24,25].
Moreover, since a neural network usually has a spatial nature due to the presence of an amount of parallel pathways
of a variety of axon sizes and lengths, it is desired to model them by introducing continuously distributed delays over
certain duration of time such that the distant past has less influence compared to the recent behavior of the state [26].
Especially, we can use integro-differential equations to describe a system possessing hereditary properties. Recently, the
stability of Cohen–Grossberg neural networks with continuously distributed delay has been studied in [27–30].
To the best of our knowledge, few authors have considered the stability and periodicity of impulsive Cohen–Gross-
berg neural networks with continuously distributed delays, which is very important in theories and applications and
also is a very challenging problem. Motivated by the above discussion, in this paper, we will investigate stability and
periodicity of impulsive Cohen–Grossberg neural networks with continuously distributed delays. Firstly, global expo-
nential stability and robust global exponential stability of the equilibrium solution are investigated by using Lyapunov
function and integro-differential inequality. Secondly, sufficient conditions are given to guarantee the existence of --
periodic solution and that all other solutions are convergent to it globally exponentially. Finally, two examples are
given to show the effectiveness of our results in this paper.
This paper is organized as follows. In Section 2, we give the description of an impulsive Cohen–Grossberg neural
network with continuously distributed delays, and sufficient conditions to guarantee the existence, uniqueness and glo-
bal exponential stability of the equilibrium solution of this neural network. In Section 3, robust global exponential sta-
bility of the equilibrium solution of the uncertain impulsive Cohen–Grossberg neural network with continuously
distributed delays is investigated. In Section 4, sufficient conditions are obtained to guarantee the existence of --peri-
odic solution and that all other solutions are convergent to it globally exponentially. In Section 5, two examples are
given to illustrate the effectiveness of the proposed results in this paper.
2. Existence and uniqueness of the equilibrium solution and global exponential stability
Consider the impulsive Cohen–Grossberg neural network with continuously distributed delays as follows:
" Z t #
X
n X
n
Dxi ðtÞ ¼ ai ðxi ðtÞÞ bi ðxi ðtÞÞ cij fj ðxj ðtÞÞuj d ij k ij ðt sÞfj ðxj ðsÞÞdsxj þ I i ; ð1Þ
j¼1 j¼1 1
where i = 1, 2, . . . , n, n P 2 is the number of neurons in this network; D denotes the distributional derivative;
x = (x1, x2, . . . , xn)T 2 Rn and xi(t) is the state variable of the ith neuron; ai() > 0 represents an amplification function;
bi() is the appropriately behaved function; (cij)nn and (dij)nn denote the connection matrix; fj() is the non-linear in-
T n
put–output activation of jth neuron; R 1 I = (I1, I2, . . . , In) 2 R denotes input vector; kernels kij:[0, +1) ? [0, +1) are
piecewise continuous and satisfy 0 k ij ðsÞ ds ¼ 1. uj and xj satisfy:
X1 X1
uj ¼ 1 þ bjk dðt tk Þ; xj ¼ 1 þ cjk dðt tk Þ; ð2Þ
k¼1 k¼1
where t1 < t2 < < tk < , limk?+1tk = +1, d(t) is the Dirac impulsive function, bjk and cjk are constants.
We assume that the following conditions are satisfied:
(A2) bi ðxÞb
xy
i ðyÞ
P ci > 0 for all x, y 2 R, x – y, i = 1, 2, . . . , n;
(A3) there exist Lj such that
jfj ðxÞ fj ðyÞj 6 Lj jx yj; 8x; y 2 R;
where Lj > 0 are constants, j = 1,2, . . . , n.
(3) z > 0 and the constant k satisfies 0 < k 6 k1 and is determined by the following inequality
Z 1
kþpþ qðsÞeks ds z < 0;
0
then,
uðtÞ 6 z ekðtt0 Þ ; t P t0 ;
provided that the initial conditions satisfy uðsÞ 6 z ekðst0 Þ ; 1 < s 6 t0 .
Now, we present the global exponential stability criteria for the impulsive Cohen–Grossberg neural network (1) with
continuously distributed delays.
Theorem 1. Assume that (A1)–(A3) hold. The impulsive Cohen–Grossberg neural network (1) with continuously distributed
delays has a unique equilibrium solution which is globally exponentially stable if the following conditions hold:
Proof. Firstly, we will prove the existence and uniqueness of equilibrium solution for the impulsive Cohen–Grossberg
neural network (1) with continuously delays. Supposing x* is the equilibrium solution of the neural network (1), then x*
satisfies:
X
n X
n Z 1
bi ðxi Þ cij fj ðxj Þuj d ij k ij ðsÞfj ðxj Þds xj þ I i ¼ 0: ð3Þ
j¼1 j¼1 0
Z.W. Ping, J.G. Lu / Chaos, Solitons and Fractals 41 (2009) 164–174 167
R1
From 0
k ij ðsÞds ¼ 1 and (3), we can obtain
Xn X
n
bi ðxi Þ cij fj ðxj Þuj d ij fj ðxj Þ xj þ I i ¼ 0; ð4Þ
j¼1 j¼1
which is equivalent to
X
n X
n X
n X
n
bi ðxi Þ cij fj ðxj Þ d ij fj ðxj Þ þ I i ¼ 0; cij fj ðxj Þuj þ d ij fj ðxj Þxj ¼ 0: ð5Þ
j¼1 j¼1 j¼1 j¼1
Similar to [24], we can conclude that the impulsive Cohen–Grossberg neural network (1) with continuously delays has a
unique equilibrium solution when the condition (i) in Theorem 1 holds.
In the following, we will prove the unique equilibrium solution x ¼ ðx1 ; x2 ; . . . ; xn ÞT is globally exponentially stable.
In order to simplify our proofs we will shift the equilibrium solution x* of the impulsive Cohen–Grossberg neural
network (1) to the origin. The transformation y i ðtÞ ¼ xi ðtÞ xi puts the impulsive Cohen–Grossberg neural network (1)
into the following form:
" Z 1 #
X
n X
n
Dy i ðtÞ ¼ ai ðxi ðtÞÞ bi ðxi ðtÞÞ bi ðxi Þ cij ðfj ðxj ðtÞÞ fj ðxj ÞÞuj d ij k ij ðsÞðfj ðxj ðt sÞÞ fj ðxj ÞÞds xj
j¼1 j¼1 0
" Z #
X
n X
n 1
¼ ai ðy i ðtÞÞ bi ðy i ðtÞÞ cij uj ðy j ðtÞÞuj d ij k ij ðsÞuj ðy j ðt sÞÞds xj ; ð6Þ
j¼1 j¼1 0
where
ai ðy i ðtÞÞ ¼ ai ðy i ðtÞ þ xi Þ; bi ðy i ðtÞÞ ¼ bi ðy i ðtÞ þ xi Þ bi ðxi Þ; uj ðy j ðtÞÞ ¼ fj ðy i ðtÞ þ xi Þ fj ðxj Þ: ð7Þ
Construct the Lyapunov function as follows:
Xn Z y i ðtÞ
sgnðsÞ
V ðtÞ ¼ ds: ð8Þ
i¼1 0 ai ðsÞ
Then,
" Z #
X
n X
n 1
jy i ðtk Þj 6 jy i ðt
k Þj þ
ai jcij bjk jLj jy j ðtk Þj þ jd ij cjk j k ij ðsÞLj jy j ðtk sÞjds ; ð12Þ
j¼1 j¼1 0
Define kV t0 k ¼ sup1<t6t0 fV ðtÞg; k/k ¼ sup1<s6t0 fkyðsÞkg: The following results come from (10) and (15). For
t 2 (t0, t1], according to (10) and Lemma 1, we have
1
V ðtÞ 6 kV t0 kekðtt0 Þ 6 k/kekðtt0 Þ ; ð16Þ
a
which leads to
1
V ðt
1 Þ 6 kV t0 ke
kðt1 t0 Þ
6 k/kekðt1 t0 Þ ;
a
and
Z 1 Z 1
a
a
V ðtþ
1Þ 6 p1 V ðt
1 Þ þ q1 ðsÞV ðt 1 sÞds 6 k/k p 1 ekðt1 t0 Þ
þ q1 ðsÞeks
ds ekðt1 t0 Þ
: ð17Þ
a 0 a2 0
So,
a
kyðtÞk 6 k/keðkqÞðtt0 Þ ¼ M 0 k/keðkqÞðtt0 Þ ; ð26Þ
a
where M 0 ¼ ða=aÞ P 1; k q > 0.
Therefore the equilibrium solution of the impulsive Cohen–Grossberg neural network (1) with continuously
distributed delays is globally exponentially stable. This completes the proof of Theorem 1. h
In this section, we will establish the robust global exponential stability criteria for the uncertain impulsive Cohen–
Grossberg neural network with continuously distributed delays.
Consider the uncertain impulsive Cohen–Grossberg neural network with continuously distributed delays as follows:
" Z t #
X
n X
n
Dxi ðtÞ ¼ ai ðxi ðtÞÞ bi ðxi ðtÞÞ cij fj ðxj ðtÞÞuj d ij k ij ðt sÞfj ðxj ðsÞÞdsxj þ I i ; ð27Þ
j¼1 j¼1 1
where cij 6 cij 6 cij , d ij 6 d ij 6 dij , the other notations and conditions are the same as those in the impulsive Cohen–
Grossberg neural networks (1).
ð0Þ ð1Þ ð0Þ ð1Þ
Let cij ¼ 12 ðcij þ cij Þ; cij ¼ 12 ðcij cij Þ; d ij ¼ 12 ðdij þ d ij Þ; d ij ¼ 12 ðdij d ij Þ.
Theorem 2. Assume that (A1)–(A3) hold. The uncertain impulsive Cohen–Grossberg neural network (27) with continuously
distributed delays has a unique equilibrium solution which is robust globally exponentially stable if the following conditions hold:
4. Periodic solution of the impulsive Cohen–Grossberg neural network with continuously distributed delays
In this section, we will discuss the periodic solution of impulsive Cohen–Grossberg neural network with continu-
ously distributed delays.
Consider the following impulsive Cohen–Grossberg neural network with continuously distributed delays:
8 R
< Dxi ðtÞ ¼ ai ðt; xi ðtÞÞ½bi ðt; xi ðtÞÞ P cij ðtÞfj ðxj ðtÞÞ P d ij ðtÞ t k ij ðt sÞfj ðxj ðsÞÞds þ I i ðtÞ; t–tk ;
n n
1
j¼1 j¼1 ð28Þ
:
xi ðtk Þ ¼ ð1 þ bik Þxi ðt
k Þ; i ¼ 1; 2; . . . n;
170 Z.W. Ping, J.G. Lu / Chaos, Solitons and Fractals 41 (2009) 164–174
where
(i) ai 2 C(R2, (0, +1)) is --periodic with respect to the first argument and there exist constants ai and
ai such that
ai 6 ai 6 ai ;
(ii) bi 2 C(R2, R) is --periodic with respect to the first argument;
(iii) cij, dij, Ii 2 C(R, R) are --periodic;
(iv) For - > 0, there exists q 2 N such that tk + - = tk+q and bi(k+q) = bik, k = 1,2, .
Assume that the initial condition is xi(t) = /i(t), t 6 0. We have the following theorem.
Theorem 3. Assume that
(i) jai ðt; uÞ ai ðt; vÞj 6 Lai ju vj; 8u; v 2 R; t 2 ½0; þ-;
(ii) oðai ðt;uÞb
ou
i ðt;uÞÞ
P cab i ; t 2 ½0; -; u 2 R;
(iii) |fj(u) P fj (v)| 6 Lj|u v|, |fj(u)| 6 Mj, "u, v 2 R; Pn
(iv) cabi
n
aj cji Li þ Lai ðcji þ dji ÞM i Lai I M
j¼1 ½ i aj Li dji > 0; 8t 2 ½0; þ1Þ;
j¼1
Then there is a unique --periodic solution for the impulsive Cohen–Grossberg neural network (28) and all other solu-
tions are convergent to it globally exponentially.
Proof. Let zi(t) = xi(t + -) xi(t), where xi(t + -) = wi (t), t 6 0.
When t – tk, calculating the upper right Dini-derivative D+zi(t) of zi(t) along the solution of system (28) and
estimating its right hand side, we have
" Z 1 #
X
n X
n
Dzi ðtÞ ¼ ai ðt; xi ðt þ -ÞÞ bi ðt; xi ðt þ -ÞÞ cij ðtÞfj ðxj ðt þ -ÞÞ d ij ðtÞ k ij ðsÞfj ðxj ðt þ - sÞÞds þ I i ðtÞ
j¼1 j¼1 0
" Z #
X
n X
n 1
þ ai ðt; xi ðtÞÞ bi ðt; xi ðtÞÞ cij ðtÞfj ðxj ðtÞÞ d ij ðtÞ k ij ðsÞfj ðxj ðt sÞÞds þ I i ðtÞ
j¼1 j¼1 0
X
n
½ai ðt; xi ðt þ -ÞÞbi ðt; xi ðt þ -ÞÞ ai ðt; xi ðtÞÞbi ðt; xi ðtÞÞ þ ai ðt; xi ðt þ -ÞÞ cij ðtÞ½fj ðxj ðt þ -ÞÞ
j¼1
X
n X
n Z 1
fi ðxj ðtÞÞ þ ½ai ðxi ðt þ -ÞÞ ai ðxi ðtÞÞ cij ðtÞfj ðxj ðtÞÞ þ ai ðt; xi ðt þ -ÞÞ d ij ðtÞ k ij ðsÞ½fj ðxj ðt þ - sÞÞ
j¼1 j¼1 0
X
n Z 1
fj ðxj ðt sÞÞds þ ½ai ðxi ðt þ -ÞÞ ai ðxi ðtÞÞ d ij k ij ðsÞfj ðxj ðt sÞÞds ½ai ðxi ðt þ -ÞÞ ai ðxi ðtÞÞI i ðtÞ
j¼1 0
¼ ai ðzi ðtÞÞ þ bi ðzi ðtÞÞ þ ci ðzi ðtÞÞ þ P i ðzi ðtÞÞ þ Qi ðzi ðtÞÞ Ri ðzi ðtÞÞ; t–tk ; ð29Þ
where
ai ðzi ðtÞÞ ¼ ai ðt; xi ðt þ -ÞÞbi ðt; xi ðt þ -ÞÞ ai ðt; xi ðtÞÞbi ðt; xi ðtÞÞ;
X n
bi ðzi ðtÞÞ ¼ ai ðt; xi ðt þ -ÞÞ cij ðtÞ½fj ðxj ðt þ -ÞÞ fj ðxj ðtÞÞ;
j¼1
X
n
ci ðzi ðtÞÞ ¼ ½ai ðxi ðt þ -ÞÞ ai ðxi ðtÞÞ cij ðtÞfj ðxj ðtÞÞ;
j¼1
X
n Z 1
P i ðzi ðtÞÞ ¼ d ij ðtÞ k ij ðsÞ½fj ðxj ðt þ - sÞÞ fj ðxj ðt sÞÞds;
j¼1 0
X
n Z 1
Qi ðzi ðtÞÞ ¼ ½ai ðxi ðt þ -ÞÞ ai ðxi ðtÞÞ d ij k ij ðsÞfj ðxj ðt sÞÞds;
j¼1 0
ð33Þ
On the other hand, by using condition (v) in Theorem 3,
V ðtk Þ 6 max j1 þ bik jV ðt
k Þ 6 V ðtk Þ: ð34Þ
16i6n
Therefore,
V ðtÞ 6 V ð0Þ; t 2 ½tk ; tkþ1 Þ; k 2 N: ð35Þ
By the definition of V(t), we have
" Z Z #
Xn X
n 1 0
V ð0Þ 6 jzi ð0Þj þ ai Lj k ij ðsÞ expfeðu þ sÞgdij jzj ðuÞjdu ds
i¼1 j¼1 0 s
" Z Z #
X
n X
n 1 0
6 jzi ð0Þj þ aj Li d ji k ji ðsÞees jzi ðuÞjdu ds
i¼1 j¼1 0 s
X
n Z 1
6 k/ wk þ aj Li d ji k ji ðsÞees dsk/ wk 6 Kk/ wk: ð36Þ
j¼1 0
P
Note that V ðtÞ P ni¼1 eet jzi ðtÞj, thus
Xn
jzi ðtÞj 6 K eet k/ wk; 8t P 0: ð37Þ
i¼1
Therefore
jxi ðt þ -Þ xi ðtÞj 6 K eet k/ wk: ð38Þ
P
In view of xi ðt þ k-Þ ¼ xi ðtÞ þ kl¼1 ½xi ðt þ l-Þ xi ðt þ ðl 1Þ-Þ, the following limit exists.
172 Z.W. Ping, J.G. Lu / Chaos, Solitons and Fractals 41 (2009) 164–174
Define
xi ðtÞ ¼ lim xi ðt þ k-Þ;
k!1
T ð39Þ
Y ðtÞ ¼ x1 ðtÞ; x2 ðtÞ; . . . ; xn ðtÞ :
Thus Y*(t) is a periodic function with periodic - for system (28). Moreover, assuming system (28) has two periodic
solutions ZðtÞ; ZðtÞ, and by a similar method used before, we can get the following result:
jxi ðt þ k-Þ xi ðt þ k-Þj 6 K eek- ; ð40Þ
which means that the limit solution is unique for system (28), and all other solutions are convergent to it globally
exponentially. h
5. Example
In order to illustrate the effectiveness of the presented results, we give the following two examples.
Example 1. Consider the following impulsive Cohen–Grossberg neural network with continuously distributed delays:
" 1
!
1
x_ 1 ðtÞ 0:2 þ 0:01 sin x1 0 3x1 ðtÞ 4 4
sinðx1 ðtÞÞu1
¼ 1 1
x_ 2 ðtÞ 0 0:2 þ 0:01 cos x2 3x2 ðtÞ 4 4
cosðx2 ðtÞÞu2
Z t ! # ð41Þ
14 kðt sÞ 18 kðt sÞ sinðx1 ðsÞÞw1 0
ds þ
1 161 kðt sÞ 121 kðt sÞ cosðx2 ðsÞÞw2 1:5p
P1
where k(t) = exp(t) is the kernel function, u1 ¼ u2 ¼ w1 ¼ w2 ¼ 1 þ k¼1 1:5dðt tk Þ, tk tk1 = 1.4. By simple calcu-
lation, from (41), we have a ¼ 0:21, a = 0.19, L1 = L2 = 1, c1 = c2 = 3, and
Example 2. Now, we consider the following impulsive Cohen–Grossberg neural network with continuously distributed
delays:
cos t 1 1
x_ 1 ðtÞ ¼ 6 x 1 ðtÞ sin t sinðx 1 ðtÞÞ þ cos t cosðx2 ðtÞÞ
1 þ x21 ðtÞ 10 12
Z 1 Z 1
1 1
sin t 2e2s sinðx1 ðt sÞÞds þ cos t 2e2s cosðx2 ðt sÞÞds þ sin tg; t–tk ;
21 0 30 0
sin t 1 1
x_ 2 ðtÞ ¼ 5 þ x 2 ðtÞ cos t sinðx 1 ðtÞÞ þ sin t cosðx2 ðtÞÞ ð42Þ
2 þ x22 ðtÞ 12 18
Z 1 Z 1
1 1
cos t 2e2s sinðx1 ðt sÞÞds þ sin t 2e2s cosðx2 ðt sÞÞds þ cos tg; t–tk ;
18 0 30 0
x1 ðtk Þ ¼ ð1 þ bik Þx1 ðt
k Þ;
x2 ðtk Þ ¼ ð1 þ bik Þx2 ðt
k Þ;
where
bik ¼ 1 þ ð1Þk for i ¼ 1; 2; k 2 N;
t1 ¼ p=2; t2 ¼ p; t2kþ1 ¼ t1 þ 2kp; t2k ¼ t2 þ 2ðk 1Þp; k 2 N:
Z.W. Ping, J.G. Lu / Chaos, Solitons and Fractals 41 (2009) 164–174 173
X
n X
n
cab
2 ½ajcj2 L2 þ La2 ðcj2 þ dj2 ÞM 2 La2 I M
2 aj L2 dj2
j¼1 j¼1
1 1 1 1 1 1 1 1
¼ 4 7 1 þ pffiffiffi þ 1 þ 6 1 þ pffiffiffi þ 1
12 2 30 30 18 2 30 30
1 1 1
pffiffiffi 1 7 1 þ 6 1 ¼ 1:8133 > 0;
2 30 30
(v) |1 + bik| 6 1, i = 1,2, k 2 N.
Therefore, according to Theorem 3, we conclude that the impulsive Cohen–Grossberg neural network (42) with
continuously distributed delays has a unique 2p-periodic solution which is globally exponentially stable.
6. Conclusions
In this paper, several classes of impulsive Cohen–Grossberg neural networks with continuously distributed delays
have been studied. Some sufficient conditions for global exponential stability and robust global exponential stability
of the equilibrium solution have been obtained by using Lyapunov function and integro-differential inequality. The suf-
ficient conditions that guarantee the existence of --periodic solution and that all other solutions are convergent to it
globally exponentially have also been obtained. Finally, two examples are given to demonstrate the effectiveness of
our results in this paper.
Acknowledgements
The authors would like to thank the Editor and the anonymous reviewers for their helpful comments and sugges-
tions. The research is partially supported by the National Nature Science Foundation of PR China under Grant
Nos.60404005 and 60744002.
References
[1] Cohen MA, Grossberg S. Absolute stability and global pattern formation and parallel memory storage by competitive neural
networks. IEEE Trans Syst Man Cybernet 1983;13:15–26.
[2] Huang CX, Huang LH. Dynamics of a class of Cohen–Grossberg neural networks with time-varying delays. Nonlinear Anal: Real
2007;8:40–52.
[3] Zhao HY, Wang L. Hopf bifurcation in Cohen–Grossberg neural network with distributed delays. Nonlinear Anal: Real
2007;8:73–89.
[4] Marcuss CM, Westervelt RM. Stability of analog neural networks with delay. Phys Rev A 1989;39(1):347–59.
[5] Roska T, Chua LO. Cellular neural networks with delay-type template elements and nonuniform grids. Int J Circuit Theory Appl
1992;20:469–81.
174 Z.W. Ping, J.G. Lu / Chaos, Solitons and Fractals 41 (2009) 164–174
[6] Hwang CC, Cheng CJ, Liao TL. Globally exponential stability of generalized Cohen–Grossberg neural networks with delays.
Phys Lett A 2003;319:157–66.
[7] Cao J, Liang JL. Boundedness and stability for Cohen–Grossberg neural network with time-varying delays. J Math Anal Appl
2004;296:665–85.
[8] Arik S, Orman Z. Global stability analysis of Cohen–Grossberg neural networks with time varying delays. Phys Lett A
2005;341:410–21.
[9] Yuan K, Cao J. An analysis of global asymptotic stability of delayed Cohen–Grossberg neural networks via nonsmooth analysis.
IEEE Trans Circuits Syst I 2005;52(9):1854–61.
[10] Guo S, Huang L. Stability Analysis of Cohen–Grossberg Neural Networks. IEEE Trans Neural Networks 2006;17(1):106–17.
[11] Song Q, Cao J. Stability analysis of Cohen–Grossberg neural network with both time-varying and continuously distributed delays.
J Comput Appl Math 2006;197:188–203.
[12] Song Q, Cao J. Robust global exponential stability of Cohen–Grossberg neural network with time-varying delays and reaction-
diffusion terms. J Franklin Inst 2006;343(7):705–19.
[13] Cao J, Li X. Stability in delayed Cohen–Grossberg neural network: LMI optimization approach. Phys D 2005;212:54–65.
[14] Wu W, Cui BT, Huang M. Global asymptotic stability of Cohen–Grossberg neural networks with constant and variable delays.
Chaos, Solitons & Fractals 2007;33:1355–61.
[15] Wu W, Cui BT, Huang M. Global asymptotic stability of delayed Cohen–Grossberg neural networks. Chaos, Solitons & Fractals
2007;34:872–7.
[16] Li Y. Existence and stability of periodic solutions for Cohen–Grossberg neural networks with multiple delays. Chaos, Solitons &
Fractals 2004;20:459–66.
[17] Tu F, Liao X. Harmless delays for global asymptotic stability of Cohen–Grossberg neural networks. Chaos, Solitons & Fractals
2005;26:927–33.
[18] Liu J. Global exponential stability of Cohen–Grossberg neural networks with time-varying delays. Chaos, Solitons & Fractals
2005;26:935–45.
[19] Xiong WL, Xu BG. Some criteria for robust stability of Cohen–Grossberg neural networks with delays. Chaos, Solitons &
Fractals 2006;36:1357–65.
[20] Li CH, Yang SY. A further analysis on harmless delays in Cohen–Grossberg neural networks. Chaos, Solitons & Fractals
2007;34:646–53.
[21] Zhou Q, Wan L, Sun J. Exponential stability of reaction–diffusion generalized Cohen–Grossberg neural networks with time-
varying delays. Chaos, Solitons & Fractals 2007;32:1713–9.
[22] Chen Z, Zhao D, Ruan J. Dynamic analysis of high-order Cohen–Grossberg neural networks with time delay. Chaos, Solitons &
Fractals 2007;32:1538–46.
[23] Liu B, Huang L. Existence and exponential stability of periodic solutions for a class of Cohen–Grossberg neural networks with
time-varying delays. Chaos, Solitons & Fractals 2007;32:617–27.
[24] Chen Z, Ruan J. Global stability analysis of impulsive Cohen–Grossberg neural networks with delay. Phys Lett A
2005;345:101–11.
[25] Bai C. Stability analysis of Cohen–Grossberg BAM neural networks with delays and impulses. Chaos, Solitons & Fractals
2008;35:263–7.
[26] Gopalsamy K, He XZ. Stability in asymmetric Hopfield nets with transmission delays. Phys D 1994;76:344–58.
[27] Li W, Sun J. Global asymptotic stability of Cohen–Grossberg neural network with continuously distributed delays. Phys Lett A
2005;342:331–40.
[28] Ji Y, Lou X, Cui B. Global output convergence of Cohen–Grossberg neural networks with both time-varying and distributed
delays. Chaos, Solitons & Fractals 2007;40:344–54.
[29] Huang T, Li C, Chen G. Stability of Cohen–Grossberg neural networks with unbounded distributed delays. Chaos, Solitons &
Fractals 2007;34:992–6.
[30] Li CH, Yang SY. Global attractivity in delayed Cohen–Grossberg neural network models. Chaos, Solitons & Fractals
2007;39:1975–87.
[31] Xu D, Zhu W, Long S. Global exponential stability of impulsive integro-differential equation. Nonlinear Anal 2006;64:2805–16.