+ All Categories
Home > Documents > Research Article Long Time Behavior for a System …downloads.hindawi.com/archive/2014/252674.pdf[]...

Research Article Long Time Behavior for a System …downloads.hindawi.com/archive/2014/252674.pdf[]...

Date post: 23-Aug-2020
Category:
Upload: others
View: 0 times
Download: 0 times
Share this document with a friend
8
Research Article Long Time Behavior for a System of Differential Equations with Non-Lipschitzian Nonlinearities Nasser-Eddine Tatar Department of Mathematics and Statistics, King Fahd University of Petroleum and Minerals, Dhahran 31261, Saudi Arabia Correspondence should be addressed to Nasser-Eddine Tatar; [email protected] Received 10 May 2014; Revised 7 September 2014; Accepted 8 September 2014; Published 14 September 2014 Academic Editor: Ozgur Kisi Copyright © 2014 Nasser-Eddine Tatar. is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. We consider a general system of nonlinear ordinary differential equations of first order. e nonlinearities involve distributed delays in addition to the states. In turn, the distributed delays involve nonlinear functions of the different variables and states. An explicit bound for solutions is obtained under some rather reasonable conditions. Several special cases of this system may be found in neural network theory. As a direct application of our result it is shown how to obtain global existence and, more importantly, convergence to zero at an exponential rate in a certain norm. All these nonlinearities (including the activation functions) may be non-Lipschitz and unbounded. 1. Introduction Of concern is the following system: () = − () () + =1 (, () , ∫ −∞ (, , ()) )+ () , (1) with continuous data () = 0 (), ∈ (−∞, 0], coefficients () 0, and inputs (), = 1,...,. e functions and are nonlinear continuous functions. is is a general nonlinear version of several systems that arise in many applications (see [19] and Section 4 below). e literature is very rich of works on the asymptotic behavior of solutions for special cases of system (1) (see for instance [1019]). Here the integral terms represent some kind of distributed delays but discrete delays may be recovered as well by considering delta Dirac distributions. Different sufficient conditions on the coefficients, the func- tions, and the kernels have been established ensuring con- vergence to equilibrium or (uniform, global, and asymptotic) stability. In applications it is important to have global asymp- totic stability at a very rapid rate like the exponential rate. Roughly speaking, it has been assumed that the coefficients () must dominate the coefficients of some “bad” similar terms that appear in the estimations. For the nonlinearities (activation functions), the first assumptions of boundedness, monotonicity, and differentiability have been all weakened to a Lipschitz condition. According to [8, 20] and other references, even this condition needs to be weakened further. Unfortunately, we can find only few papers on continuous but not Lipschitz continuous activation functions. Assumptions like partially Lipschitz and linear growth, -inverse H¨ older continuous or inverse Lipschitz, non-Lipschitz but bounded were used (see [16, 21, 22]). For H¨ older continuous activation functions we refer the reader to [23], where exponential stability was proved under some boundedness and monotonicity conditions on the activation functions and the coefficients form a Lyapunov diagonally stable matrix (see also [24, 25] for other results without these conditions). ere are, however, a good number of papers dealing with discontinuous activation functions under certain stronger conditions like -Matrix, the LMI condition (linear matrix inequality) and some extra conditions on the matrices and growth conditions on the activation functions (see [20, 2637]). Global asymptotic stability of periodic solutions have been investigated, for instance, in [38, 39]. Hindawi Publishing Corporation Advances in Artificial Neural Systems Volume 2014, Article ID 252674, 7 pages http://dx.doi.org/10.1155/2014/252674
Transcript
Page 1: Research Article Long Time Behavior for a System …downloads.hindawi.com/archive/2014/252674.pdf[] J. Cao, K. Yuan, and H.-X. Li, Global asymptotical stability of recurrent neural

Research ArticleLong Time Behavior for a System of Differential Equations withNon-Lipschitzian Nonlinearities

Nasser-Eddine Tatar

Department of Mathematics and Statistics King Fahd University of Petroleum and Minerals Dhahran 31261 Saudi Arabia

Correspondence should be addressed to Nasser-Eddine Tatar tatarnkfupmedusa

Received 10 May 2014 Revised 7 September 2014 Accepted 8 September 2014 Published 14 September 2014

Academic Editor Ozgur Kisi

Copyright copy 2014 Nasser-Eddine TatarThis is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

We consider a general system of nonlinear ordinary differential equations of first orderThe nonlinearities involve distributed delaysin addition to the states In turn the distributed delays involve nonlinear functions of the different variables and states An explicitbound for solutions is obtained under some rather reasonable conditions Several special cases of this systemmay be found in neuralnetwork theory As a direct application of our result it is shown how to obtain global existence and more importantly convergenceto zero at an exponential rate in a certain norm All these nonlinearities (including the activation functions) may be non-Lipschitzand unbounded

1 Introduction

Of concern is the following system

1199091015840

119894(119905) = minus119886

119894(119905) 119909119894(119905)

+

119898

sum

119895=1

119891119894119895(119905 119909119895(119905) int

119905

minusinfin

119870119894119895(119905 119904 119909

119895(119904)) 119889119904) + 119888

119894(119905)

(1)

with continuous data 119909119895(119905) = 119909

0119895(119905) 119905 isin (minusinfin 0] coefficients

119886119894(119905) ge 0 and inputs 119888

119894(119905) 119894 = 1 119898 The functions

119891119894119895and 119870

119894119895are nonlinear continuous functions This is a

general nonlinear version of several systems that arise inmany applications (see [1ndash9] and Section 4 below)

The literature is very rich of works on the asymptoticbehavior of solutions for special cases of system (1) (seefor instance [10ndash19]) Here the integral terms representsome kind of distributed delays but discrete delays may berecovered as well by considering delta Dirac distributionsDifferent sufficient conditions on the coefficients the func-tions and the kernels have been established ensuring con-vergence to equilibrium or (uniform global and asymptotic)stability In applications it is important to have global asymp-totic stability at a very rapid rate like the exponential rate

Roughly speaking it has been assumed that the coefficients119886119894(119905) must dominate the coefficients of some ldquobadrdquo similar

terms that appear in the estimations For the nonlinearities(activation functions) the first assumptions of boundednessmonotonicity and differentiability have been all weakenedto a Lipschitz condition According to [8 20] and otherreferences even this condition needs to be weakened furtherUnfortunately we can find only few papers on continuous butnot Lipschitz continuous activation functions Assumptionslike partially Lipschitz and linear growth 120572-inverse Holdercontinuous or inverse Lipschitz non-Lipschitz but boundedwere used (see [16 21 22])

For Holder continuous activation functions we referthe reader to [23] where exponential stability was provedunder some boundedness and monotonicity conditions onthe activation functions and the coefficients form a Lyapunovdiagonally stable matrix (see also [24 25] for other resultswithout these conditions)

There are however a good number of papers dealing withdiscontinuous activation functions under certain strongerconditions like 119872-Matrix the LMI condition (linear matrixinequality) and some extra conditions on the matricesand growth conditions on the activation functions (see[20 26ndash37]) Global asymptotic stability of periodic solutionshave been investigated for instance in [38 39]

Hindawi Publishing CorporationAdvances in Artificial Neural SystemsVolume 2014 Article ID 252674 7 pageshttpdxdoiorg1011552014252674

2 Advances in Artificial Neural Systems

Here we assume that the functions 119891119894119895and 119870

119894119895are (or

bounded by) continuous monotone nondecreasing functionsthat are not necessarily Lipschitz continuous and they maybe unbounded (like power type functions with powers biggerthan one) We prove that for sufficiently small initial datasolutions decay to zero exponentially

The local existence and global existence are standard seethe Gronwall-type Lemma 1 below and the estimation in ourtheorem However the uniqueness of the equilibrium is notan issue here (even in case of constant coefficients) as we areconcerned with convergence to zero rather than stability ofequilibrium

After the Preliminaries section where we present ourmain hypotheses and the main lemma used in our proofwe state and prove the convergence result in Section 3 Thesection is ended by some corollaries and important remarksIn the last section we give an application where this type ofsystems (or special cases of it) appears in real world problems

2 Preliminaries

Our first hypothesis (H1) is

10038161003816100381610038161003816100381610038161003816

119891119894119895(119905 119909119895(119905) int

119905

minusinfin

119870119894119895(119905 119904 119909

119895(119904)) 119889119904)

10038161003816100381610038161003816100381610038161003816

le 119887119894119895(119905)

10038161003816100381610038161003816119909119895(119905)

10038161003816100381610038161003816

120572119894119895

(int

119905

minusinfin

119897119894119895(119905 minus 119904) 120595

119894119895(

10038161003816100381610038161003816119909119895(119904)

10038161003816100381610038161003816) 119889119904)

120573119894119895

119894 119895 = 1 119898

(2)

where 119887119894119895are nonnegative continuous functions 119897

119894119895are non-

negative continuously differentiable functions 120595119894119895are non-

negative nondecreasing continuous functions and 120572119894119895 120573119894119895ge

0 119894119895 = 1 119898 The interesting cases are when 120572119894119895and 120573

119894119895

are all nonzeroLet 119868 sub R and let 119892

1 1198922 119868 rarr R 0 We write 119892

1prop 1198922

if 11989221198921is nondecreasing in 119868 This ordering as well as the

monotonicity condition may be dropped as is mentioned inRemark 8 below

Lemma 1 (see [40]) Let 119886(119905) be a positive continuous functionin 119869 = [120572 120573) 119896

119895(119905) 119895 = 1 119899 nonnegative continuous

functions for 120572 le 119905 lt 120573 119892119895(119906) 119895 = 1 119899 nondecreasing

continuous functions in R+ with 119892

119895(119906) gt 0 for 119906 gt 0 and 119906(119905)

a nonnegative continuous functions in 119869 If 1198921prop 1198922prop sdot sdot sdot prop

119892119899in (0infin) then the inequality

119906 (119905) le 119886 (119905) +

119899

sum

119895=1

int

119905

120572

119896119895(119904) 119892119895(119906 (119904)) 119889119904 119905 isin 119869 (3)

implies that

119906 (119905) le 120596119899(119905) 120572 le 119905 lt 120573

0 (4)

where 1205960(119905) = sup

0le119904le119905119886(119904)

120596119895(119905) = 119866

minus1

119895[119866119895(120596119895minus1

(119905)) + int

119905

0

119896119895(119904) 119889119904] 119895 = 1 119899

119866119895(119906) = int

119906

119906119895

119889119909

119892119895(119909)

119906 gt 0 (119906119895gt 0 119895 = 1 119899)

(5)

and 1205730is chosen so that the functions 120596

119895(119905) 119895 = 1 119899 are

defined for 120572 le 119905 lt 1205730

In our case we will need the following notation andhypotheses

(H2) Assume that 120595119894119895(119906) gt 0 for 119906 gt 0 and the set of

functions 119906(119905)120572119894119895+120573119894119895 120595119894119895(119906(119905)) may be ordered as ℎ

1prop ℎ2prop

sdot sdot sdot prop ℎ119899(after relabelling) Their corresponding coefficients

119887119894119895(119905) = exp[int119905

0119886(120590)119889120590]119887

119894119895(119905) (119886(119905) = min

1le119894le119898119886119894(119905)) and 119897

119894119895(0)

will be renamed 120582119896 119896 = 1 119899

We define 119909(119905) = sum119898

119894=1|119909119894(119905)| 119905 gt 0 119909

0(119905) = sum

119898

119894=1|1199090119894(119905)|

119905 le 0

119888 (119905) = int

119905

0

exp [int119904

0

119886 (120590) 119889120590]

119898

sum

119894=1

1003816100381610038161003816119888119894(119904)1003816100381610038161003816119889119904 119905 gt 0

1205960(119905) = 119909

0(0) +

119898

sum

119894119895=1

int

0

minusinfin

119897119894119895(minus120590) 120595

119894119895(1199090(120590)) 119889120590 + 119888 (119905)

120596119895(119905) = 119867

minus1

119895[119867119895(120596119895minus1

(119905)) + int

119905

120572

120582119895(119904) 119889119904] 119895 = 1 119899

119867119895(119906) = int

119906

119906119895

119889119909

ℎ119895(119909)

119906 gt 0 (119906119895gt 0 119895 = 1 119899)

0(119905) = 120596

0(0) +

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816int

0

minus119904

120595119894119895(1199060(120590)) 119889120590 119889119904

1199060(120590) =

119898

sum

119894119895=1

int

120590

minusinfin

10038161003816100381610038161003816119897119894119895(120590 minus 120591)

10038161003816100381610038161003816120595119894119895(1199090(120591)) 119889120591 120590 lt 0

119895(119905) = 119867

minus1

119895[119867119895(119895minus1

(119905)) + int

119905

0

120582119895(119904) 119889119904] 119895 = 1 119899

(6)

where 120582119895are the relabelled coefficients corresponding to119887

119894119895(119905)

and 119897119894119895(0) + int

infin

0|1198971015840

119894119895(120590)|119889120590

3 Exponential Convergence

In this section it is proved that solutions converge to zero inan exponentialmanner provided that the initial data are smallenough

Advances in Artificial Neural Systems 3

Theorem 2 Assume that the hypotheses (H1) and (H2) holdand int0minusinfin

119897119894119895(minus120590)120595

119894119895(1199090(120590))119889120590 lt infin 119894 119895 = 1 119898 Then (a)

if 1198971015840119894119895(119905) le 0 119894 119895 = 1 119898 there exists 120573

0gt 0 such that

119909 (119905) le 120596119899(119905) exp [minusint

119905

0

119886 (119904) 119889119904] 0 le 119905 lt 1205730 (7)

(b) If 1198971015840119894119895(119905) 119894 119895 = 1 119898 are of arbitrary signs 1198971015840

119894119895(119905) are

summable and the integral term in 0(119905) is convergent then

there exists a 1205731gt 0 such that the conclusion in (a) is valid on

0 le 119905 lt 1205731with

119899instead of 120596

119899

Proof It is easy to see from (1) and the assumption (H1) thatfor 119905 gt 0 and 119894 = 1 119898 we have

119863+ 1003816100381610038161003816119909119894(119905)1003816100381610038161003816le minus119886119894(119905)

1003816100381610038161003816119909119894(119905)1003816100381610038161003816

+

119898

sum

119895=1

10038161003816100381610038161003816100381610038161003816

119891119894119895(119905 119909119895(119905) int

119905

minusinfin

119870119894119895(119905 119904 119909

119895(119904)) 119889119904)

10038161003816100381610038161003816100381610038161003816

+ 119888119894(119905)

(8)

or for 119905 gt 0

119863+119909 (119905) le minusmin

1le119894le119898

119886119894(119905) 119909 (119905)

+

119898

sum

119894119895=1

119887119894119895(119905)

10038161003816100381610038161003816119909119895(119905)

10038161003816100381610038161003816

120572119894119895

times (int

119905

minusinfin

119897119894119895(119905 minus 119904)120595

119894119895(

10038161003816100381610038161003816119909119895(119904)

10038161003816100381610038161003816) 119889119904)

120573119894119895

+

119898

sum

119894=1

1003816100381610038161003816119888119894(119905)1003816100381610038161003816

(9)

where119863+ denotes the right Dini derivative Hence

119863+119909 (119905)

le minus119886 (119905) 119909 (119905)

+

119898

sum

119894119895=1

119887119894119895(119905) |119909 (119905)|

120572119894119895(int

119905

minusinfin

119897119894119895(119905 minus 119904) 120595

119894119895(119909 (119904)) 119889119904)

120573119894119895

+

119898

sum

119894=1

1003816100381610038161003816119888119894(119905)1003816100381610038161003816 119905 gt 0

(10)

and consequently

119863+119909 (119905) exp [int

119905

0

119886 (119904) 119889119904]

le exp [int119905

0

119886 (119904) 119889119904]

119898

sum

119894119895=1

119887119894119895(119905) |119909(119905)|

120572119894119895

times (int

119905

minusinfin

119897119894119895(119905 minus 119904)120595

119894119895(119909(119904)) 119889119904)

120573119894119895

+ exp [int119905

0

119886 (119904) 119889119904]

119898

sum

119894=1

1003816100381610038161003816119888119894(119905)1003816100381610038161003816

119905 gt 0

(11)

Thus (by a comparison theorem in [41])

119909 (119905) le 119909 (0) + 119888 (119905)

+

119898

sum

119895=1

int

119905

0

119898

sum

119894=1

119887119894119895(119904) |119909 (119904)|

120572119894119895

times(int

119904

minusinfin

119897119894119895(119904 minus 120590) 120595

119894119895(119909 (120590)) 119889120590)

120573119894119895

119889119904

119905 gt 0

(12)

where

119909 (119905) = 119909 (119905) exp [int119905

0

119886 (119904) 119889119904] (13)

Let 119910(119905) denote the right hand side of (12) Clearly 119909(119905) le

119910(119905) 119905 gt 0 and for 119905 gt 0

119863+119910 (119905) = 119863

+119888 (119905)

+

119898

sum

119894119895=1

119887119894119895(119905) |119909 (119905)|

120572119894119895

times (int

119905

minusinfin

119897119894119895(119905 minus 120590)120595

119894119895(119909(120590)) 119889120590)

120573119894119895

(14)

We designate by 119911119894119895(119905) the integral term in (14) that is

119911119894119895(119905) = int

119905

minusinfin

119897119894119895(119905 minus 120590) 120595

119894119895(119909 (120590)) 119889120590 (15)

and 119911(119905) = sum119898

119894119895=1119911119894119895(119905) A differentiation of 119911(119905) gives

1199111015840

(119905) =

119898

sum

119894119895=1

119897119894119895(0) 120595119894119895(119909 (119905))

+

119898

sum

119894119895=1

int

119905

minusinfin

1198971015840

119894119895(119905 minus 120590) 120595

119894119895(119909 (120590)) 119889120590

(16)

(a) Consider 1198971015840119894119895(119905) le 0 119894 119895 = 1 119898

In this situation (of fading memory) we see from (14) and(16) that if 119906(119905) = 119910(119905) + 119911(119905) then

119863+119906 (119905) le 119863

+119888 (119905)

+

119898

sum

119894119895=1

[119887119894119895(119905) (119906 (119905))

120572119894119895+120573119894119895+ 119897119894119895(0) 120595119894119895(119906 (119905))]

119905 gt 0

(17)

4 Advances in Artificial Neural Systems

Therefore119906 (119905) le 119906 (0) + 119888 (119905)

+

119898

sum

119894119895=1

int

119905

0

[119887119894119895(119904) (119906 (119904))

120572119894119895+120573119894119895+ 119897119894119895(0) 120595119894119895(119906 (119904))] 119889119904

119905 gt 0

(18)

where 119906(0) = 119909(0) + sum119898

119894119895=1int

0

minusinfin119897119894119895(minus120590)120595

119894119895(1199090(120590))119889120590 Now we

can apply Lemma 1 to obtain119909 (119905) le 119906 (119905) le 120596

119899(119905) 0 le 119905 lt 120573

0(19)

with 1205960(119905) = 119906(0) + 119888(119905) and 120596

119899(119905) is as in the ldquoPreliminariesrdquo

section(b) Consider 1198971015840

119894119895(119905) 119894 119895 = 1 119898 of arbitrary signs

From expressions (14) and (16) we derive that

119863+119906 (119905) le 119863

+119888 (119905)

+

119898

sum

119894119895=1

[119887119894119895(119905) (119906(119905))

120572119894119895+120573119894119895+ 119897119894119895(0) 120595119894119895(119906 (119905))]

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(120590)

10038161003816100381610038161003816120595119894119895(119906 (119905 minus 120590)) 119889120590 119905 gt 0

(20)

The derivative of the auxiliary function

(119905) = 119906 (119905) +

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816int

119905

119905minus119904

120595119894119895(119906 (120590)) 119889120590 119889119904

119905 ge 0

(21)

is equal to (with the help of (20) and (21))

119863+ (119905) = 119863

+119906 (119905)

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816[120595119894119895(119906 (119905)) minus 120595

119894119895(119906 (119905 minus 119904))] 119889120590 119889119904

le 119863+119888 (119905) +

119898

sum

119894119895=1

[119887119894119895(119905) (119906 (119905))

120572119894119895+120573119894119895+ 119897119894119895(0) 120595119894119895(119906 (119905))]

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(120590)

10038161003816100381610038161003816120595119894119895(119906 (119905 minus 120590)) 119889120590

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816[120595119894119895(119906 (119905)) minus 120595

119894119895(119906 (119905 minus 119904))] 119889119904

le 119863+119888 (119905)

+

119898

sum

119894119895=1

119887119894119895(119905) ( (119905))

120572119894119895+120573119894119895

+[119897119894119895(0) + int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816119889119904]120595119894119895( (119905))

119905 gt 0

(22)

Therefore

(119905) le (0) + 119888 (119905)

+

119898

sum

119894119895=1

int

119905

0

119887119894119895(119904) ( (119904))

120572119894119895+120573119894119895

+[119897119894119895(0) + int

infin

0

100381610038161003816100381610038161198971015840

119894119895(120590)

10038161003816100381610038161003816119889120590]120595

119894119895( (119904)) 119889119904

(23)

with

(0) = 119909 (0) +

119898

sum

119894119895=1

int

0

minusinfin

119897119894119895(minus120590) 120595

119894119895(1199090(120590)) 119889120590

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816int

0

minus119904

120595119894119895(1199060(120590)) 119889120590 119889119904

1199060(120590) = 119911 (120590) =

119898

sum

119894119895=1

119911119894119895(120590)

=

119898

sum

119894119895=1

int

120590

minusinfin

119897119894119895(120590 minus 120591) 120595

119894119895(1199090(120591)) 119889120591 120590 lt 0

(24)

Applying Lemma 1 to (23) we obtain

119909 (119905) le (119905) le 119899(119905) 0 le 119905 lt 120573

1(25)

and hence

119909 (119905) le 119899(119905) 0 le 119905 lt 120573

1 (26)

where 0(119905) = (0) and

119895(119905) = 119867

minus1

119895[119867119895(119895minus1

(119905)) + int

119905

0

120582119895(119904) 119889119904]

119895 = 1 119899

(27)

and 1205730is chosen so that the functions

119895(119905) 119895 = 1 119899 are

defined for 0 le 119905 lt 1205731

Corollary 3 If in addition to the hypotheses of the theoremwe assume that

int

infin

0

120594119896(119904) 119889119904 le int

infin

120596119896minus1

119889119911

ℎ119896(119911)

119896 = 1 119899 120594119896(119904) = 120582

119896(119904)

120582119896(119904)

(28)

then we have global existence of solutions

Corollary 4 If in addition to the hypotheses of the theoremwe assume that120596

119899(119905) (119899(119905)) grows up at themost polynomially

(or just slower than exp[int1199050119886(119904)119889119904]) then solutions decay at an

exponential rate if int1199050119886(119904)119889119904 rarr infin as 119905 rarr infin

Corollary 5 In addition to the hypotheses of the theoremassume that 1198971015840

119894119895(119905) le 119871

119894119895119897119894119895(119905) 119894 119895 = 1 119898 for some positive

Advances in Artificial Neural Systems 5

constants 119871119894119895and 120595

119894119895(119905) are in the class H (that is 120595

119894119895(120572119906) le

120585119894119895120595119894119895(119906) 120572 gt 0 119906 gt 0 119894 119895 = 1 119898) Then solutions are

bounded by a function of the form exp[minus(int1199050119886(119904)119889119904minus119871119905)] where

119897 = max119871119894119895 119894 119895 = 1 119898

Remark 6 We have assumed that 120572119894119895and 120573

119894119895are greater than

one but the case when they are smaller than one may betreated similarlyWhen their sum is smaller than one we haveglobal existence without adding any extra condition

Remark 7 The decay rate obtained in Corollary 5 is to becompared with the one in the theorem (case (b)) It appearsthat the estimation in Corollary 5 holds for more generalinitial data (not as small as the ones in case (b)) Howeverthe decay rate is smaller than the one in (b) besides assumingthat int1199050119886(119904)119889119904 minus 119871119905 rarr infin as 119905 rarr infin

Remark 8 If we consider the following new functions thenthe monotonicity condition and the order imposed in thetheorem may be dropped

1206011(119905) = max

0le119904le119905

1198921(119904)

120601119896(119905) = max

0le119904le119905

119892119896(119904)

120601119896minus1

(119904)

120601119896minus1

(119905)

(29)

and 120595(119905) = 120601119896(119905)120601119896minus1

(119905)

4 Application

(Artificial)Neural networks are built in an attempt to performdifferent tasks just as the nervous system Typically a neuralnetwork consists of several layers (input layer hidden layersand output layer) Each layer contains one or more cells(neurons) with many connections between them The cellsin one layer receive inputs from the previous layer makesome transformations and send the results to the cells of thesubsequent layer

One may encounter neural networks in many fields suchas control patternmatching settlement of structures classifi-cation of soil supply chain management engineering designmarket segmentation product analysis market developmentforecasting signature verification bond rating recognitionof diseases robust pattern detection text mining price fore-cast botanical classification and scheduling optimization

Neural networks not only can perform many of the tasksa traditional computer can do but also excel in for instanceclassifying incomplete or noisy data predicting future eventsand generalizing

The system (1) is a general version of simpler systems thatappear in neural network theory [1ndash9] like

1199091015840

119894(119905) = minus119886

119894119909119894(119905) +

119898

sum

119895=1

119891119894119895(119909119895(119905)) + 119888

119894(119905) (30)

or

1199091015840

119894(119905) = minus119886

119894119909119894(119905)

+

119898

sum

119895=1

int

119905

minusinfin

119897119894119895(119905 minus 119904) 119891

119894119895(119909119895(119904)) 119889119904 + 119888

119894(119905)

(31)

It is well established by now that (for constant coefficientsand constant 119888

119894(119905)) solutions converge in an exponential

manner to the equilibrium Notice that zero in our case isnot an equilibrium This equilibrium exists and is unique incase of Lipschitz continuity of the activation functions Inour case the system is much more general and the activationfunctions as well as the nonlinearities are not necessarily Lip-schitz continuous However in case of Lipschitz continuityand existence of a unique equilibrium we expect to haveexponential stability using the standard techniques at leastwhen we start away from zero

For the system

1199091015840

119894(119905) = minus119886

119894119909119894(119905)

+

119898

sum

119895=1

119887119894119895

10038161003816100381610038161003816119909119895(119905)

10038161003816100381610038161003816

120572119894119895

(int

119905

minusinfin

119897119894119895(119905 minus 119904) 120595

119894119895(

10038161003816100381610038161003816119909119895(119904)

10038161003816100381610038161003816) 119889119904)

120573119894119895

+ 119888119894(119905)

(32)

(where 120595119894119895

may be taken as power functions see alsoCorollary 5) our theorem gives sufficient conditions guaran-teeing the estimation

119909 (119905) le 120596119899(119905) exp [minusint

119905

0

119886 (119904) 119889119904] 0 le 119905 lt 1205730 (33)

Then Corollaries 3 and 4 provide practical situations wherewe have global existence and decay to zero at an exponentialrate

Conflict of Interests

The author declares that there is no conflict of interestsregarding the publication of this paper

Acknowledgment

The author is grateful for the financial support and thefacilities provided by King Fahd University of Petroleum andMinerals through Grant no IN111052

References

[1] J Cao K Yuan and H-X Li ldquoGlobal asymptotical stabilityof recurrent neural networks with multiple discrete delays anddistributed delaysrdquo IEEE Transactions on Neural Networks vol17 no 6 pp 1646ndash1651 2006

[2] B Crespi ldquoStorage capacity of non-monotonic neuronsrdquoNeuralNetworks vol 12 no 10 pp 1377ndash1389 1999

6 Advances in Artificial Neural Systems

[3] G de Sandre M Forti P Nistri and A Premoli ldquoDynamicalanalysis of full-range cellular neural networks by exploiting dif-ferential variational inequalitiesrdquo IEEE Transactions on Circuitsand Systems I Regular Papers vol 54 no 8 pp 1736ndash1749 2007

[4] C Feng and R Plamondon ldquoOn the stability analysis of delayedneural networks systemsrdquo Neural Networks vol 14 no 9 pp1181ndash1188 2001

[5] J J Hopfield ldquoNeural networks and physical systems withemergent collective computational abilitiesrdquo Proceedings of theNational Academy of Sciences of the United States of Americavol 79 no 8 pp 2554ndash2558 1982

[6] J J Hopfield and D W Tank ldquoComputing with neural circuitsa modelrdquo Science vol 233 no 4764 pp 625ndash633 1986

[7] J I Inoue ldquoRetrieval phase diagrams of non-monotonic Hop-field networksrdquo Journal of Physics A Mathematical and Generalvol 29 no 16 pp 4815ndash4826 1996

[8] B Kosko Neural Network and Fuzzy SystemmdashA DynamicalSystem Approach toMachine Intelligence Prentice-Hall of IndiaNew Delhi India 1991

[9] H-F Yanai and S-I Amari ldquoAuto-associative memory withtwo-stage dynamics of nonmonotonic neuronsrdquo IEEE Transac-tions on Neural Networks vol 7 no 4 pp 803ndash815 1996

[10] X Liu and N Jiang ldquoRobust stability analysis of generalizedneural networks with multiple discrete delays and multipledistributed delaysrdquo Neurocomputing vol 72 no 7ndash9 pp 1789ndash1796 2009

[11] SMohamadKGopalsamy andHAkca ldquoExponential stabilityof artificial neural networks with distributed delays and largeimpulsesrdquo Nonlinear Analysis Real World Applications vol 9no 3 pp 872ndash888 2008

[12] J Park ldquoOn global stability criterion for neural networks withdiscrete and distributed delaysrdquo Chaos Solitons and Fractalsvol 30 no 4 pp 897ndash902 2006

[13] J Park ldquoOn global stability criterion of neural networks withcontinuously distributed delaysrdquo Chaos Solitons and Fractalsvol 37 no 2 pp 444ndash449 2008

[14] Z Qiang M A Run-Nian and X Jin ldquoGlobal exponentialconvergence analysis of Hopfield neural networks with con-tinuously distributed delaysrdquo Communications in TheoreticalPhysics vol 39 no 3 pp 381ndash384 2003

[15] Y Wang W Xiong Q Zhou B Xiao and Y Yu ldquoGlobal expo-nential stability of cellular neural networks with continuouslydistributed delays and impulsesrdquo Physics Letters A vol 350 no1-2 pp 89ndash95 2006

[16] H Wu ldquoGlobal exponential stability of Hopfield neural net-works with delays and inverse Lipschitz neuron activationsrdquoNonlinear Analysis Real World Applications vol 10 no 4 pp2297ndash2306 2009

[17] Q Zhang X P Wei and J Xu ldquoGlobal exponential stability ofHopfield neural networkswith continuously distributed delaysrdquoPhysics Letters A vol 315 no 6 pp 431ndash436 2003

[18] H Zhao ldquoGlobal asymptotic stability of Hopfield neural net-work involving distributed delaysrdquo Neural Networks vol 17 no1 pp 47ndash53 2004

[19] J Zhou S Li and Z Yang ldquoGlobal exponential stability ofHopfield neural networks with distributed delaysrdquo AppliedMathematical Modelling Simulation and Computation for Engi-neering and Environmental Systems vol 33 no 3 pp 1513ndash15202009

[20] R Gavalda and H T Siegelmann ldquoDiscontinuities in recurrentneural networksrdquo Neural Computation vol 11 no 3 pp 715ndash745 1999

[21] H Wu F Tao L Qin R Shi and L He ldquoRobust exponentialstability for interval neural networks with delays and non-Lipschitz activation functionsrdquoNonlinearDynamics vol 66 no4 pp 479ndash487 2011

[22] H Wu and X Xue ldquoStability analysis for neural networks withinverse Lipschitzian neuron activations and impulsesrdquo AppliedMathematical Modelling vol 32 no 11 pp 2347ndash2359 2008

[23] M Forti M Grazzini P Nistri and L Pancioni ldquoGeneralizedLyapunov approach for convergence of neural networks withdiscontinuous or non-LIPschitz activationsrdquo Physica D Nonlin-ear Phenomena vol 214 no 1 pp 88ndash99 2006

[24] N-E Tatar ldquoHopfield neural networks with unbounded mono-tone activation functionsrdquoAdvances inArtificial Neural Systemsvol 2012 Article ID 571358 5 pages 2012

[25] N-E Tatar ldquoControl of systems with Holder continuousfunctions in the distributed delaysrdquo Carpathian Journal ofMathematics vol 30 no 1 pp 123ndash128 2014

[26] G Bao and Z Zeng ldquoAnalysis and design of associative mem-ories based on recurrent neural network with discontinuousactivation functionsrdquoNeurocomputing vol 77 no 1 pp 101ndash1072012

[27] M Forti and P Nistri ldquoGlobal convergence of neural networkswith discontinuous neuron activationsrdquo IEEE Transactions onCircuits and Systems I Fundamental Theory and Applicationsvol 50 no 11 pp 1421ndash1435 2003

[28] YHuang H Zhang and ZWang ldquoDynamical stability analysisof multiple equilibrium points in time-varying delayed recur-rent neural networks with discontinuous activation functionsrdquoNeurocomputing vol 91 pp 21ndash28 2012

[29] L Li and L Huang ldquoDynamical behaviors of a class of recur-rent neural networks with discontinuous neuron activationsrdquoApplied Mathematical Modelling vol 33 no 12 pp 4326ndash43362009

[30] L Li and L Huang ldquoGlobal asymptotic stability of delayedneural networks with discontinuous neuron activationsrdquo Neu-rocomputing vol 72 no 16-18 pp 3726ndash3733 2009

[31] Y Li and H Wu ldquoGlobal stability analysis for periodic solutionin discontinuous neural networks with nonlinear growth acti-vationsrdquo Advances in Difference Equations vol 2009 Article ID798685 14 pages 2009

[32] X Liu and J Cao ldquoRobust state estimation for neural networkswith discontinuous activationsrdquo IEEE Transactions on SystemsMan and Cybernetics B Cybernetics vol 40 no 6 pp 1425ndash1437 2010

[33] J Liu X Liu and W-C Xie ldquoGlobal convergence of neuralnetworks with mixed time-varying delays and discontinuousneuron activationsrdquo Information Sciences vol 183 pp 92ndash1052012

[34] S Qin and X Xue ldquoGlobal exponential stability and global con-vergence in finite time of neural networks with discontinuousactivationsrdquoNeural Processing Letters vol 29 no 3 pp 189ndash2042009

[35] J Wang L Huang and Z Guo ldquoGlobal asymptotic stabilityof neural networks with discontinuous activationsrdquo NeuralNetworks vol 22 no 7 pp 931ndash937 2009

[36] Z Wang L Huang Y Zuo and L Zhang ldquoGlobal robuststability of time-delay systems with discontinuous activationfunctions under polytopic parameter uncertaintiesrdquo Bulletin ofthe KoreanMathematical Society vol 47 no 1 pp 89ndash102 2010

[37] H Wu ldquoGlobal stability analysis of a general class of discontin-uous neural networks with linear growth activation functionsrdquoInformation Sciences vol 179 no 19 pp 3432ndash3441 2009

Advances in Artificial Neural Systems 7

[38] Z Cai and L Huang ldquoExistence and global asymptotic stabilityof periodic solution for discrete and distributed time-varyingdelayed neural networks with discontinuous activationsrdquo Neu-rocomputing vol 74 no 17 pp 3170ndash3179 2011

[39] D Papini and V Taddei ldquoGlobal exponential stability of theperiodic solution of a delayed neural network with discontinu-ous activationsrdquo Physics Letters A vol 343 no 1ndash3 pp 117ndash1282005

[40] M Pinto ldquoIntegral inequalities of Bihari-type and applicationsrdquoFunkcialaj Ekvacioj vol 33 no 3 pp 387ndash403 1990

[41] V Lakshmikhantam and S Leela Differential and IntegralInequalities Theory and Applications vol 55-I of Mathematicsin Sciences and Engineering Edited by Bellman R AcademicPress New York NY USA 1969

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Distributed Sensor Networks

International Journal of

Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014

International Journal of

ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ArtificialNeural Systems

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Computational Intelligence and Neuroscience

Industrial EngineeringJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Page 2: Research Article Long Time Behavior for a System …downloads.hindawi.com/archive/2014/252674.pdf[] J. Cao, K. Yuan, and H.-X. Li, Global asymptotical stability of recurrent neural

2 Advances in Artificial Neural Systems

Here we assume that the functions 119891119894119895and 119870

119894119895are (or

bounded by) continuous monotone nondecreasing functionsthat are not necessarily Lipschitz continuous and they maybe unbounded (like power type functions with powers biggerthan one) We prove that for sufficiently small initial datasolutions decay to zero exponentially

The local existence and global existence are standard seethe Gronwall-type Lemma 1 below and the estimation in ourtheorem However the uniqueness of the equilibrium is notan issue here (even in case of constant coefficients) as we areconcerned with convergence to zero rather than stability ofequilibrium

After the Preliminaries section where we present ourmain hypotheses and the main lemma used in our proofwe state and prove the convergence result in Section 3 Thesection is ended by some corollaries and important remarksIn the last section we give an application where this type ofsystems (or special cases of it) appears in real world problems

2 Preliminaries

Our first hypothesis (H1) is

10038161003816100381610038161003816100381610038161003816

119891119894119895(119905 119909119895(119905) int

119905

minusinfin

119870119894119895(119905 119904 119909

119895(119904)) 119889119904)

10038161003816100381610038161003816100381610038161003816

le 119887119894119895(119905)

10038161003816100381610038161003816119909119895(119905)

10038161003816100381610038161003816

120572119894119895

(int

119905

minusinfin

119897119894119895(119905 minus 119904) 120595

119894119895(

10038161003816100381610038161003816119909119895(119904)

10038161003816100381610038161003816) 119889119904)

120573119894119895

119894 119895 = 1 119898

(2)

where 119887119894119895are nonnegative continuous functions 119897

119894119895are non-

negative continuously differentiable functions 120595119894119895are non-

negative nondecreasing continuous functions and 120572119894119895 120573119894119895ge

0 119894119895 = 1 119898 The interesting cases are when 120572119894119895and 120573

119894119895

are all nonzeroLet 119868 sub R and let 119892

1 1198922 119868 rarr R 0 We write 119892

1prop 1198922

if 11989221198921is nondecreasing in 119868 This ordering as well as the

monotonicity condition may be dropped as is mentioned inRemark 8 below

Lemma 1 (see [40]) Let 119886(119905) be a positive continuous functionin 119869 = [120572 120573) 119896

119895(119905) 119895 = 1 119899 nonnegative continuous

functions for 120572 le 119905 lt 120573 119892119895(119906) 119895 = 1 119899 nondecreasing

continuous functions in R+ with 119892

119895(119906) gt 0 for 119906 gt 0 and 119906(119905)

a nonnegative continuous functions in 119869 If 1198921prop 1198922prop sdot sdot sdot prop

119892119899in (0infin) then the inequality

119906 (119905) le 119886 (119905) +

119899

sum

119895=1

int

119905

120572

119896119895(119904) 119892119895(119906 (119904)) 119889119904 119905 isin 119869 (3)

implies that

119906 (119905) le 120596119899(119905) 120572 le 119905 lt 120573

0 (4)

where 1205960(119905) = sup

0le119904le119905119886(119904)

120596119895(119905) = 119866

minus1

119895[119866119895(120596119895minus1

(119905)) + int

119905

0

119896119895(119904) 119889119904] 119895 = 1 119899

119866119895(119906) = int

119906

119906119895

119889119909

119892119895(119909)

119906 gt 0 (119906119895gt 0 119895 = 1 119899)

(5)

and 1205730is chosen so that the functions 120596

119895(119905) 119895 = 1 119899 are

defined for 120572 le 119905 lt 1205730

In our case we will need the following notation andhypotheses

(H2) Assume that 120595119894119895(119906) gt 0 for 119906 gt 0 and the set of

functions 119906(119905)120572119894119895+120573119894119895 120595119894119895(119906(119905)) may be ordered as ℎ

1prop ℎ2prop

sdot sdot sdot prop ℎ119899(after relabelling) Their corresponding coefficients

119887119894119895(119905) = exp[int119905

0119886(120590)119889120590]119887

119894119895(119905) (119886(119905) = min

1le119894le119898119886119894(119905)) and 119897

119894119895(0)

will be renamed 120582119896 119896 = 1 119899

We define 119909(119905) = sum119898

119894=1|119909119894(119905)| 119905 gt 0 119909

0(119905) = sum

119898

119894=1|1199090119894(119905)|

119905 le 0

119888 (119905) = int

119905

0

exp [int119904

0

119886 (120590) 119889120590]

119898

sum

119894=1

1003816100381610038161003816119888119894(119904)1003816100381610038161003816119889119904 119905 gt 0

1205960(119905) = 119909

0(0) +

119898

sum

119894119895=1

int

0

minusinfin

119897119894119895(minus120590) 120595

119894119895(1199090(120590)) 119889120590 + 119888 (119905)

120596119895(119905) = 119867

minus1

119895[119867119895(120596119895minus1

(119905)) + int

119905

120572

120582119895(119904) 119889119904] 119895 = 1 119899

119867119895(119906) = int

119906

119906119895

119889119909

ℎ119895(119909)

119906 gt 0 (119906119895gt 0 119895 = 1 119899)

0(119905) = 120596

0(0) +

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816int

0

minus119904

120595119894119895(1199060(120590)) 119889120590 119889119904

1199060(120590) =

119898

sum

119894119895=1

int

120590

minusinfin

10038161003816100381610038161003816119897119894119895(120590 minus 120591)

10038161003816100381610038161003816120595119894119895(1199090(120591)) 119889120591 120590 lt 0

119895(119905) = 119867

minus1

119895[119867119895(119895minus1

(119905)) + int

119905

0

120582119895(119904) 119889119904] 119895 = 1 119899

(6)

where 120582119895are the relabelled coefficients corresponding to119887

119894119895(119905)

and 119897119894119895(0) + int

infin

0|1198971015840

119894119895(120590)|119889120590

3 Exponential Convergence

In this section it is proved that solutions converge to zero inan exponentialmanner provided that the initial data are smallenough

Advances in Artificial Neural Systems 3

Theorem 2 Assume that the hypotheses (H1) and (H2) holdand int0minusinfin

119897119894119895(minus120590)120595

119894119895(1199090(120590))119889120590 lt infin 119894 119895 = 1 119898 Then (a)

if 1198971015840119894119895(119905) le 0 119894 119895 = 1 119898 there exists 120573

0gt 0 such that

119909 (119905) le 120596119899(119905) exp [minusint

119905

0

119886 (119904) 119889119904] 0 le 119905 lt 1205730 (7)

(b) If 1198971015840119894119895(119905) 119894 119895 = 1 119898 are of arbitrary signs 1198971015840

119894119895(119905) are

summable and the integral term in 0(119905) is convergent then

there exists a 1205731gt 0 such that the conclusion in (a) is valid on

0 le 119905 lt 1205731with

119899instead of 120596

119899

Proof It is easy to see from (1) and the assumption (H1) thatfor 119905 gt 0 and 119894 = 1 119898 we have

119863+ 1003816100381610038161003816119909119894(119905)1003816100381610038161003816le minus119886119894(119905)

1003816100381610038161003816119909119894(119905)1003816100381610038161003816

+

119898

sum

119895=1

10038161003816100381610038161003816100381610038161003816

119891119894119895(119905 119909119895(119905) int

119905

minusinfin

119870119894119895(119905 119904 119909

119895(119904)) 119889119904)

10038161003816100381610038161003816100381610038161003816

+ 119888119894(119905)

(8)

or for 119905 gt 0

119863+119909 (119905) le minusmin

1le119894le119898

119886119894(119905) 119909 (119905)

+

119898

sum

119894119895=1

119887119894119895(119905)

10038161003816100381610038161003816119909119895(119905)

10038161003816100381610038161003816

120572119894119895

times (int

119905

minusinfin

119897119894119895(119905 minus 119904)120595

119894119895(

10038161003816100381610038161003816119909119895(119904)

10038161003816100381610038161003816) 119889119904)

120573119894119895

+

119898

sum

119894=1

1003816100381610038161003816119888119894(119905)1003816100381610038161003816

(9)

where119863+ denotes the right Dini derivative Hence

119863+119909 (119905)

le minus119886 (119905) 119909 (119905)

+

119898

sum

119894119895=1

119887119894119895(119905) |119909 (119905)|

120572119894119895(int

119905

minusinfin

119897119894119895(119905 minus 119904) 120595

119894119895(119909 (119904)) 119889119904)

120573119894119895

+

119898

sum

119894=1

1003816100381610038161003816119888119894(119905)1003816100381610038161003816 119905 gt 0

(10)

and consequently

119863+119909 (119905) exp [int

119905

0

119886 (119904) 119889119904]

le exp [int119905

0

119886 (119904) 119889119904]

119898

sum

119894119895=1

119887119894119895(119905) |119909(119905)|

120572119894119895

times (int

119905

minusinfin

119897119894119895(119905 minus 119904)120595

119894119895(119909(119904)) 119889119904)

120573119894119895

+ exp [int119905

0

119886 (119904) 119889119904]

119898

sum

119894=1

1003816100381610038161003816119888119894(119905)1003816100381610038161003816

119905 gt 0

(11)

Thus (by a comparison theorem in [41])

119909 (119905) le 119909 (0) + 119888 (119905)

+

119898

sum

119895=1

int

119905

0

119898

sum

119894=1

119887119894119895(119904) |119909 (119904)|

120572119894119895

times(int

119904

minusinfin

119897119894119895(119904 minus 120590) 120595

119894119895(119909 (120590)) 119889120590)

120573119894119895

119889119904

119905 gt 0

(12)

where

119909 (119905) = 119909 (119905) exp [int119905

0

119886 (119904) 119889119904] (13)

Let 119910(119905) denote the right hand side of (12) Clearly 119909(119905) le

119910(119905) 119905 gt 0 and for 119905 gt 0

119863+119910 (119905) = 119863

+119888 (119905)

+

119898

sum

119894119895=1

119887119894119895(119905) |119909 (119905)|

120572119894119895

times (int

119905

minusinfin

119897119894119895(119905 minus 120590)120595

119894119895(119909(120590)) 119889120590)

120573119894119895

(14)

We designate by 119911119894119895(119905) the integral term in (14) that is

119911119894119895(119905) = int

119905

minusinfin

119897119894119895(119905 minus 120590) 120595

119894119895(119909 (120590)) 119889120590 (15)

and 119911(119905) = sum119898

119894119895=1119911119894119895(119905) A differentiation of 119911(119905) gives

1199111015840

(119905) =

119898

sum

119894119895=1

119897119894119895(0) 120595119894119895(119909 (119905))

+

119898

sum

119894119895=1

int

119905

minusinfin

1198971015840

119894119895(119905 minus 120590) 120595

119894119895(119909 (120590)) 119889120590

(16)

(a) Consider 1198971015840119894119895(119905) le 0 119894 119895 = 1 119898

In this situation (of fading memory) we see from (14) and(16) that if 119906(119905) = 119910(119905) + 119911(119905) then

119863+119906 (119905) le 119863

+119888 (119905)

+

119898

sum

119894119895=1

[119887119894119895(119905) (119906 (119905))

120572119894119895+120573119894119895+ 119897119894119895(0) 120595119894119895(119906 (119905))]

119905 gt 0

(17)

4 Advances in Artificial Neural Systems

Therefore119906 (119905) le 119906 (0) + 119888 (119905)

+

119898

sum

119894119895=1

int

119905

0

[119887119894119895(119904) (119906 (119904))

120572119894119895+120573119894119895+ 119897119894119895(0) 120595119894119895(119906 (119904))] 119889119904

119905 gt 0

(18)

where 119906(0) = 119909(0) + sum119898

119894119895=1int

0

minusinfin119897119894119895(minus120590)120595

119894119895(1199090(120590))119889120590 Now we

can apply Lemma 1 to obtain119909 (119905) le 119906 (119905) le 120596

119899(119905) 0 le 119905 lt 120573

0(19)

with 1205960(119905) = 119906(0) + 119888(119905) and 120596

119899(119905) is as in the ldquoPreliminariesrdquo

section(b) Consider 1198971015840

119894119895(119905) 119894 119895 = 1 119898 of arbitrary signs

From expressions (14) and (16) we derive that

119863+119906 (119905) le 119863

+119888 (119905)

+

119898

sum

119894119895=1

[119887119894119895(119905) (119906(119905))

120572119894119895+120573119894119895+ 119897119894119895(0) 120595119894119895(119906 (119905))]

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(120590)

10038161003816100381610038161003816120595119894119895(119906 (119905 minus 120590)) 119889120590 119905 gt 0

(20)

The derivative of the auxiliary function

(119905) = 119906 (119905) +

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816int

119905

119905minus119904

120595119894119895(119906 (120590)) 119889120590 119889119904

119905 ge 0

(21)

is equal to (with the help of (20) and (21))

119863+ (119905) = 119863

+119906 (119905)

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816[120595119894119895(119906 (119905)) minus 120595

119894119895(119906 (119905 minus 119904))] 119889120590 119889119904

le 119863+119888 (119905) +

119898

sum

119894119895=1

[119887119894119895(119905) (119906 (119905))

120572119894119895+120573119894119895+ 119897119894119895(0) 120595119894119895(119906 (119905))]

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(120590)

10038161003816100381610038161003816120595119894119895(119906 (119905 minus 120590)) 119889120590

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816[120595119894119895(119906 (119905)) minus 120595

119894119895(119906 (119905 minus 119904))] 119889119904

le 119863+119888 (119905)

+

119898

sum

119894119895=1

119887119894119895(119905) ( (119905))

120572119894119895+120573119894119895

+[119897119894119895(0) + int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816119889119904]120595119894119895( (119905))

119905 gt 0

(22)

Therefore

(119905) le (0) + 119888 (119905)

+

119898

sum

119894119895=1

int

119905

0

119887119894119895(119904) ( (119904))

120572119894119895+120573119894119895

+[119897119894119895(0) + int

infin

0

100381610038161003816100381610038161198971015840

119894119895(120590)

10038161003816100381610038161003816119889120590]120595

119894119895( (119904)) 119889119904

(23)

with

(0) = 119909 (0) +

119898

sum

119894119895=1

int

0

minusinfin

119897119894119895(minus120590) 120595

119894119895(1199090(120590)) 119889120590

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816int

0

minus119904

120595119894119895(1199060(120590)) 119889120590 119889119904

1199060(120590) = 119911 (120590) =

119898

sum

119894119895=1

119911119894119895(120590)

=

119898

sum

119894119895=1

int

120590

minusinfin

119897119894119895(120590 minus 120591) 120595

119894119895(1199090(120591)) 119889120591 120590 lt 0

(24)

Applying Lemma 1 to (23) we obtain

119909 (119905) le (119905) le 119899(119905) 0 le 119905 lt 120573

1(25)

and hence

119909 (119905) le 119899(119905) 0 le 119905 lt 120573

1 (26)

where 0(119905) = (0) and

119895(119905) = 119867

minus1

119895[119867119895(119895minus1

(119905)) + int

119905

0

120582119895(119904) 119889119904]

119895 = 1 119899

(27)

and 1205730is chosen so that the functions

119895(119905) 119895 = 1 119899 are

defined for 0 le 119905 lt 1205731

Corollary 3 If in addition to the hypotheses of the theoremwe assume that

int

infin

0

120594119896(119904) 119889119904 le int

infin

120596119896minus1

119889119911

ℎ119896(119911)

119896 = 1 119899 120594119896(119904) = 120582

119896(119904)

120582119896(119904)

(28)

then we have global existence of solutions

Corollary 4 If in addition to the hypotheses of the theoremwe assume that120596

119899(119905) (119899(119905)) grows up at themost polynomially

(or just slower than exp[int1199050119886(119904)119889119904]) then solutions decay at an

exponential rate if int1199050119886(119904)119889119904 rarr infin as 119905 rarr infin

Corollary 5 In addition to the hypotheses of the theoremassume that 1198971015840

119894119895(119905) le 119871

119894119895119897119894119895(119905) 119894 119895 = 1 119898 for some positive

Advances in Artificial Neural Systems 5

constants 119871119894119895and 120595

119894119895(119905) are in the class H (that is 120595

119894119895(120572119906) le

120585119894119895120595119894119895(119906) 120572 gt 0 119906 gt 0 119894 119895 = 1 119898) Then solutions are

bounded by a function of the form exp[minus(int1199050119886(119904)119889119904minus119871119905)] where

119897 = max119871119894119895 119894 119895 = 1 119898

Remark 6 We have assumed that 120572119894119895and 120573

119894119895are greater than

one but the case when they are smaller than one may betreated similarlyWhen their sum is smaller than one we haveglobal existence without adding any extra condition

Remark 7 The decay rate obtained in Corollary 5 is to becompared with the one in the theorem (case (b)) It appearsthat the estimation in Corollary 5 holds for more generalinitial data (not as small as the ones in case (b)) Howeverthe decay rate is smaller than the one in (b) besides assumingthat int1199050119886(119904)119889119904 minus 119871119905 rarr infin as 119905 rarr infin

Remark 8 If we consider the following new functions thenthe monotonicity condition and the order imposed in thetheorem may be dropped

1206011(119905) = max

0le119904le119905

1198921(119904)

120601119896(119905) = max

0le119904le119905

119892119896(119904)

120601119896minus1

(119904)

120601119896minus1

(119905)

(29)

and 120595(119905) = 120601119896(119905)120601119896minus1

(119905)

4 Application

(Artificial)Neural networks are built in an attempt to performdifferent tasks just as the nervous system Typically a neuralnetwork consists of several layers (input layer hidden layersand output layer) Each layer contains one or more cells(neurons) with many connections between them The cellsin one layer receive inputs from the previous layer makesome transformations and send the results to the cells of thesubsequent layer

One may encounter neural networks in many fields suchas control patternmatching settlement of structures classifi-cation of soil supply chain management engineering designmarket segmentation product analysis market developmentforecasting signature verification bond rating recognitionof diseases robust pattern detection text mining price fore-cast botanical classification and scheduling optimization

Neural networks not only can perform many of the tasksa traditional computer can do but also excel in for instanceclassifying incomplete or noisy data predicting future eventsand generalizing

The system (1) is a general version of simpler systems thatappear in neural network theory [1ndash9] like

1199091015840

119894(119905) = minus119886

119894119909119894(119905) +

119898

sum

119895=1

119891119894119895(119909119895(119905)) + 119888

119894(119905) (30)

or

1199091015840

119894(119905) = minus119886

119894119909119894(119905)

+

119898

sum

119895=1

int

119905

minusinfin

119897119894119895(119905 minus 119904) 119891

119894119895(119909119895(119904)) 119889119904 + 119888

119894(119905)

(31)

It is well established by now that (for constant coefficientsand constant 119888

119894(119905)) solutions converge in an exponential

manner to the equilibrium Notice that zero in our case isnot an equilibrium This equilibrium exists and is unique incase of Lipschitz continuity of the activation functions Inour case the system is much more general and the activationfunctions as well as the nonlinearities are not necessarily Lip-schitz continuous However in case of Lipschitz continuityand existence of a unique equilibrium we expect to haveexponential stability using the standard techniques at leastwhen we start away from zero

For the system

1199091015840

119894(119905) = minus119886

119894119909119894(119905)

+

119898

sum

119895=1

119887119894119895

10038161003816100381610038161003816119909119895(119905)

10038161003816100381610038161003816

120572119894119895

(int

119905

minusinfin

119897119894119895(119905 minus 119904) 120595

119894119895(

10038161003816100381610038161003816119909119895(119904)

10038161003816100381610038161003816) 119889119904)

120573119894119895

+ 119888119894(119905)

(32)

(where 120595119894119895

may be taken as power functions see alsoCorollary 5) our theorem gives sufficient conditions guaran-teeing the estimation

119909 (119905) le 120596119899(119905) exp [minusint

119905

0

119886 (119904) 119889119904] 0 le 119905 lt 1205730 (33)

Then Corollaries 3 and 4 provide practical situations wherewe have global existence and decay to zero at an exponentialrate

Conflict of Interests

The author declares that there is no conflict of interestsregarding the publication of this paper

Acknowledgment

The author is grateful for the financial support and thefacilities provided by King Fahd University of Petroleum andMinerals through Grant no IN111052

References

[1] J Cao K Yuan and H-X Li ldquoGlobal asymptotical stabilityof recurrent neural networks with multiple discrete delays anddistributed delaysrdquo IEEE Transactions on Neural Networks vol17 no 6 pp 1646ndash1651 2006

[2] B Crespi ldquoStorage capacity of non-monotonic neuronsrdquoNeuralNetworks vol 12 no 10 pp 1377ndash1389 1999

6 Advances in Artificial Neural Systems

[3] G de Sandre M Forti P Nistri and A Premoli ldquoDynamicalanalysis of full-range cellular neural networks by exploiting dif-ferential variational inequalitiesrdquo IEEE Transactions on Circuitsand Systems I Regular Papers vol 54 no 8 pp 1736ndash1749 2007

[4] C Feng and R Plamondon ldquoOn the stability analysis of delayedneural networks systemsrdquo Neural Networks vol 14 no 9 pp1181ndash1188 2001

[5] J J Hopfield ldquoNeural networks and physical systems withemergent collective computational abilitiesrdquo Proceedings of theNational Academy of Sciences of the United States of Americavol 79 no 8 pp 2554ndash2558 1982

[6] J J Hopfield and D W Tank ldquoComputing with neural circuitsa modelrdquo Science vol 233 no 4764 pp 625ndash633 1986

[7] J I Inoue ldquoRetrieval phase diagrams of non-monotonic Hop-field networksrdquo Journal of Physics A Mathematical and Generalvol 29 no 16 pp 4815ndash4826 1996

[8] B Kosko Neural Network and Fuzzy SystemmdashA DynamicalSystem Approach toMachine Intelligence Prentice-Hall of IndiaNew Delhi India 1991

[9] H-F Yanai and S-I Amari ldquoAuto-associative memory withtwo-stage dynamics of nonmonotonic neuronsrdquo IEEE Transac-tions on Neural Networks vol 7 no 4 pp 803ndash815 1996

[10] X Liu and N Jiang ldquoRobust stability analysis of generalizedneural networks with multiple discrete delays and multipledistributed delaysrdquo Neurocomputing vol 72 no 7ndash9 pp 1789ndash1796 2009

[11] SMohamadKGopalsamy andHAkca ldquoExponential stabilityof artificial neural networks with distributed delays and largeimpulsesrdquo Nonlinear Analysis Real World Applications vol 9no 3 pp 872ndash888 2008

[12] J Park ldquoOn global stability criterion for neural networks withdiscrete and distributed delaysrdquo Chaos Solitons and Fractalsvol 30 no 4 pp 897ndash902 2006

[13] J Park ldquoOn global stability criterion of neural networks withcontinuously distributed delaysrdquo Chaos Solitons and Fractalsvol 37 no 2 pp 444ndash449 2008

[14] Z Qiang M A Run-Nian and X Jin ldquoGlobal exponentialconvergence analysis of Hopfield neural networks with con-tinuously distributed delaysrdquo Communications in TheoreticalPhysics vol 39 no 3 pp 381ndash384 2003

[15] Y Wang W Xiong Q Zhou B Xiao and Y Yu ldquoGlobal expo-nential stability of cellular neural networks with continuouslydistributed delays and impulsesrdquo Physics Letters A vol 350 no1-2 pp 89ndash95 2006

[16] H Wu ldquoGlobal exponential stability of Hopfield neural net-works with delays and inverse Lipschitz neuron activationsrdquoNonlinear Analysis Real World Applications vol 10 no 4 pp2297ndash2306 2009

[17] Q Zhang X P Wei and J Xu ldquoGlobal exponential stability ofHopfield neural networkswith continuously distributed delaysrdquoPhysics Letters A vol 315 no 6 pp 431ndash436 2003

[18] H Zhao ldquoGlobal asymptotic stability of Hopfield neural net-work involving distributed delaysrdquo Neural Networks vol 17 no1 pp 47ndash53 2004

[19] J Zhou S Li and Z Yang ldquoGlobal exponential stability ofHopfield neural networks with distributed delaysrdquo AppliedMathematical Modelling Simulation and Computation for Engi-neering and Environmental Systems vol 33 no 3 pp 1513ndash15202009

[20] R Gavalda and H T Siegelmann ldquoDiscontinuities in recurrentneural networksrdquo Neural Computation vol 11 no 3 pp 715ndash745 1999

[21] H Wu F Tao L Qin R Shi and L He ldquoRobust exponentialstability for interval neural networks with delays and non-Lipschitz activation functionsrdquoNonlinearDynamics vol 66 no4 pp 479ndash487 2011

[22] H Wu and X Xue ldquoStability analysis for neural networks withinverse Lipschitzian neuron activations and impulsesrdquo AppliedMathematical Modelling vol 32 no 11 pp 2347ndash2359 2008

[23] M Forti M Grazzini P Nistri and L Pancioni ldquoGeneralizedLyapunov approach for convergence of neural networks withdiscontinuous or non-LIPschitz activationsrdquo Physica D Nonlin-ear Phenomena vol 214 no 1 pp 88ndash99 2006

[24] N-E Tatar ldquoHopfield neural networks with unbounded mono-tone activation functionsrdquoAdvances inArtificial Neural Systemsvol 2012 Article ID 571358 5 pages 2012

[25] N-E Tatar ldquoControl of systems with Holder continuousfunctions in the distributed delaysrdquo Carpathian Journal ofMathematics vol 30 no 1 pp 123ndash128 2014

[26] G Bao and Z Zeng ldquoAnalysis and design of associative mem-ories based on recurrent neural network with discontinuousactivation functionsrdquoNeurocomputing vol 77 no 1 pp 101ndash1072012

[27] M Forti and P Nistri ldquoGlobal convergence of neural networkswith discontinuous neuron activationsrdquo IEEE Transactions onCircuits and Systems I Fundamental Theory and Applicationsvol 50 no 11 pp 1421ndash1435 2003

[28] YHuang H Zhang and ZWang ldquoDynamical stability analysisof multiple equilibrium points in time-varying delayed recur-rent neural networks with discontinuous activation functionsrdquoNeurocomputing vol 91 pp 21ndash28 2012

[29] L Li and L Huang ldquoDynamical behaviors of a class of recur-rent neural networks with discontinuous neuron activationsrdquoApplied Mathematical Modelling vol 33 no 12 pp 4326ndash43362009

[30] L Li and L Huang ldquoGlobal asymptotic stability of delayedneural networks with discontinuous neuron activationsrdquo Neu-rocomputing vol 72 no 16-18 pp 3726ndash3733 2009

[31] Y Li and H Wu ldquoGlobal stability analysis for periodic solutionin discontinuous neural networks with nonlinear growth acti-vationsrdquo Advances in Difference Equations vol 2009 Article ID798685 14 pages 2009

[32] X Liu and J Cao ldquoRobust state estimation for neural networkswith discontinuous activationsrdquo IEEE Transactions on SystemsMan and Cybernetics B Cybernetics vol 40 no 6 pp 1425ndash1437 2010

[33] J Liu X Liu and W-C Xie ldquoGlobal convergence of neuralnetworks with mixed time-varying delays and discontinuousneuron activationsrdquo Information Sciences vol 183 pp 92ndash1052012

[34] S Qin and X Xue ldquoGlobal exponential stability and global con-vergence in finite time of neural networks with discontinuousactivationsrdquoNeural Processing Letters vol 29 no 3 pp 189ndash2042009

[35] J Wang L Huang and Z Guo ldquoGlobal asymptotic stabilityof neural networks with discontinuous activationsrdquo NeuralNetworks vol 22 no 7 pp 931ndash937 2009

[36] Z Wang L Huang Y Zuo and L Zhang ldquoGlobal robuststability of time-delay systems with discontinuous activationfunctions under polytopic parameter uncertaintiesrdquo Bulletin ofthe KoreanMathematical Society vol 47 no 1 pp 89ndash102 2010

[37] H Wu ldquoGlobal stability analysis of a general class of discontin-uous neural networks with linear growth activation functionsrdquoInformation Sciences vol 179 no 19 pp 3432ndash3441 2009

Advances in Artificial Neural Systems 7

[38] Z Cai and L Huang ldquoExistence and global asymptotic stabilityof periodic solution for discrete and distributed time-varyingdelayed neural networks with discontinuous activationsrdquo Neu-rocomputing vol 74 no 17 pp 3170ndash3179 2011

[39] D Papini and V Taddei ldquoGlobal exponential stability of theperiodic solution of a delayed neural network with discontinu-ous activationsrdquo Physics Letters A vol 343 no 1ndash3 pp 117ndash1282005

[40] M Pinto ldquoIntegral inequalities of Bihari-type and applicationsrdquoFunkcialaj Ekvacioj vol 33 no 3 pp 387ndash403 1990

[41] V Lakshmikhantam and S Leela Differential and IntegralInequalities Theory and Applications vol 55-I of Mathematicsin Sciences and Engineering Edited by Bellman R AcademicPress New York NY USA 1969

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Distributed Sensor Networks

International Journal of

Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014

International Journal of

ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ArtificialNeural Systems

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Computational Intelligence and Neuroscience

Industrial EngineeringJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Page 3: Research Article Long Time Behavior for a System …downloads.hindawi.com/archive/2014/252674.pdf[] J. Cao, K. Yuan, and H.-X. Li, Global asymptotical stability of recurrent neural

Advances in Artificial Neural Systems 3

Theorem 2 Assume that the hypotheses (H1) and (H2) holdand int0minusinfin

119897119894119895(minus120590)120595

119894119895(1199090(120590))119889120590 lt infin 119894 119895 = 1 119898 Then (a)

if 1198971015840119894119895(119905) le 0 119894 119895 = 1 119898 there exists 120573

0gt 0 such that

119909 (119905) le 120596119899(119905) exp [minusint

119905

0

119886 (119904) 119889119904] 0 le 119905 lt 1205730 (7)

(b) If 1198971015840119894119895(119905) 119894 119895 = 1 119898 are of arbitrary signs 1198971015840

119894119895(119905) are

summable and the integral term in 0(119905) is convergent then

there exists a 1205731gt 0 such that the conclusion in (a) is valid on

0 le 119905 lt 1205731with

119899instead of 120596

119899

Proof It is easy to see from (1) and the assumption (H1) thatfor 119905 gt 0 and 119894 = 1 119898 we have

119863+ 1003816100381610038161003816119909119894(119905)1003816100381610038161003816le minus119886119894(119905)

1003816100381610038161003816119909119894(119905)1003816100381610038161003816

+

119898

sum

119895=1

10038161003816100381610038161003816100381610038161003816

119891119894119895(119905 119909119895(119905) int

119905

minusinfin

119870119894119895(119905 119904 119909

119895(119904)) 119889119904)

10038161003816100381610038161003816100381610038161003816

+ 119888119894(119905)

(8)

or for 119905 gt 0

119863+119909 (119905) le minusmin

1le119894le119898

119886119894(119905) 119909 (119905)

+

119898

sum

119894119895=1

119887119894119895(119905)

10038161003816100381610038161003816119909119895(119905)

10038161003816100381610038161003816

120572119894119895

times (int

119905

minusinfin

119897119894119895(119905 minus 119904)120595

119894119895(

10038161003816100381610038161003816119909119895(119904)

10038161003816100381610038161003816) 119889119904)

120573119894119895

+

119898

sum

119894=1

1003816100381610038161003816119888119894(119905)1003816100381610038161003816

(9)

where119863+ denotes the right Dini derivative Hence

119863+119909 (119905)

le minus119886 (119905) 119909 (119905)

+

119898

sum

119894119895=1

119887119894119895(119905) |119909 (119905)|

120572119894119895(int

119905

minusinfin

119897119894119895(119905 minus 119904) 120595

119894119895(119909 (119904)) 119889119904)

120573119894119895

+

119898

sum

119894=1

1003816100381610038161003816119888119894(119905)1003816100381610038161003816 119905 gt 0

(10)

and consequently

119863+119909 (119905) exp [int

119905

0

119886 (119904) 119889119904]

le exp [int119905

0

119886 (119904) 119889119904]

119898

sum

119894119895=1

119887119894119895(119905) |119909(119905)|

120572119894119895

times (int

119905

minusinfin

119897119894119895(119905 minus 119904)120595

119894119895(119909(119904)) 119889119904)

120573119894119895

+ exp [int119905

0

119886 (119904) 119889119904]

119898

sum

119894=1

1003816100381610038161003816119888119894(119905)1003816100381610038161003816

119905 gt 0

(11)

Thus (by a comparison theorem in [41])

119909 (119905) le 119909 (0) + 119888 (119905)

+

119898

sum

119895=1

int

119905

0

119898

sum

119894=1

119887119894119895(119904) |119909 (119904)|

120572119894119895

times(int

119904

minusinfin

119897119894119895(119904 minus 120590) 120595

119894119895(119909 (120590)) 119889120590)

120573119894119895

119889119904

119905 gt 0

(12)

where

119909 (119905) = 119909 (119905) exp [int119905

0

119886 (119904) 119889119904] (13)

Let 119910(119905) denote the right hand side of (12) Clearly 119909(119905) le

119910(119905) 119905 gt 0 and for 119905 gt 0

119863+119910 (119905) = 119863

+119888 (119905)

+

119898

sum

119894119895=1

119887119894119895(119905) |119909 (119905)|

120572119894119895

times (int

119905

minusinfin

119897119894119895(119905 minus 120590)120595

119894119895(119909(120590)) 119889120590)

120573119894119895

(14)

We designate by 119911119894119895(119905) the integral term in (14) that is

119911119894119895(119905) = int

119905

minusinfin

119897119894119895(119905 minus 120590) 120595

119894119895(119909 (120590)) 119889120590 (15)

and 119911(119905) = sum119898

119894119895=1119911119894119895(119905) A differentiation of 119911(119905) gives

1199111015840

(119905) =

119898

sum

119894119895=1

119897119894119895(0) 120595119894119895(119909 (119905))

+

119898

sum

119894119895=1

int

119905

minusinfin

1198971015840

119894119895(119905 minus 120590) 120595

119894119895(119909 (120590)) 119889120590

(16)

(a) Consider 1198971015840119894119895(119905) le 0 119894 119895 = 1 119898

In this situation (of fading memory) we see from (14) and(16) that if 119906(119905) = 119910(119905) + 119911(119905) then

119863+119906 (119905) le 119863

+119888 (119905)

+

119898

sum

119894119895=1

[119887119894119895(119905) (119906 (119905))

120572119894119895+120573119894119895+ 119897119894119895(0) 120595119894119895(119906 (119905))]

119905 gt 0

(17)

4 Advances in Artificial Neural Systems

Therefore119906 (119905) le 119906 (0) + 119888 (119905)

+

119898

sum

119894119895=1

int

119905

0

[119887119894119895(119904) (119906 (119904))

120572119894119895+120573119894119895+ 119897119894119895(0) 120595119894119895(119906 (119904))] 119889119904

119905 gt 0

(18)

where 119906(0) = 119909(0) + sum119898

119894119895=1int

0

minusinfin119897119894119895(minus120590)120595

119894119895(1199090(120590))119889120590 Now we

can apply Lemma 1 to obtain119909 (119905) le 119906 (119905) le 120596

119899(119905) 0 le 119905 lt 120573

0(19)

with 1205960(119905) = 119906(0) + 119888(119905) and 120596

119899(119905) is as in the ldquoPreliminariesrdquo

section(b) Consider 1198971015840

119894119895(119905) 119894 119895 = 1 119898 of arbitrary signs

From expressions (14) and (16) we derive that

119863+119906 (119905) le 119863

+119888 (119905)

+

119898

sum

119894119895=1

[119887119894119895(119905) (119906(119905))

120572119894119895+120573119894119895+ 119897119894119895(0) 120595119894119895(119906 (119905))]

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(120590)

10038161003816100381610038161003816120595119894119895(119906 (119905 minus 120590)) 119889120590 119905 gt 0

(20)

The derivative of the auxiliary function

(119905) = 119906 (119905) +

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816int

119905

119905minus119904

120595119894119895(119906 (120590)) 119889120590 119889119904

119905 ge 0

(21)

is equal to (with the help of (20) and (21))

119863+ (119905) = 119863

+119906 (119905)

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816[120595119894119895(119906 (119905)) minus 120595

119894119895(119906 (119905 minus 119904))] 119889120590 119889119904

le 119863+119888 (119905) +

119898

sum

119894119895=1

[119887119894119895(119905) (119906 (119905))

120572119894119895+120573119894119895+ 119897119894119895(0) 120595119894119895(119906 (119905))]

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(120590)

10038161003816100381610038161003816120595119894119895(119906 (119905 minus 120590)) 119889120590

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816[120595119894119895(119906 (119905)) minus 120595

119894119895(119906 (119905 minus 119904))] 119889119904

le 119863+119888 (119905)

+

119898

sum

119894119895=1

119887119894119895(119905) ( (119905))

120572119894119895+120573119894119895

+[119897119894119895(0) + int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816119889119904]120595119894119895( (119905))

119905 gt 0

(22)

Therefore

(119905) le (0) + 119888 (119905)

+

119898

sum

119894119895=1

int

119905

0

119887119894119895(119904) ( (119904))

120572119894119895+120573119894119895

+[119897119894119895(0) + int

infin

0

100381610038161003816100381610038161198971015840

119894119895(120590)

10038161003816100381610038161003816119889120590]120595

119894119895( (119904)) 119889119904

(23)

with

(0) = 119909 (0) +

119898

sum

119894119895=1

int

0

minusinfin

119897119894119895(minus120590) 120595

119894119895(1199090(120590)) 119889120590

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816int

0

minus119904

120595119894119895(1199060(120590)) 119889120590 119889119904

1199060(120590) = 119911 (120590) =

119898

sum

119894119895=1

119911119894119895(120590)

=

119898

sum

119894119895=1

int

120590

minusinfin

119897119894119895(120590 minus 120591) 120595

119894119895(1199090(120591)) 119889120591 120590 lt 0

(24)

Applying Lemma 1 to (23) we obtain

119909 (119905) le (119905) le 119899(119905) 0 le 119905 lt 120573

1(25)

and hence

119909 (119905) le 119899(119905) 0 le 119905 lt 120573

1 (26)

where 0(119905) = (0) and

119895(119905) = 119867

minus1

119895[119867119895(119895minus1

(119905)) + int

119905

0

120582119895(119904) 119889119904]

119895 = 1 119899

(27)

and 1205730is chosen so that the functions

119895(119905) 119895 = 1 119899 are

defined for 0 le 119905 lt 1205731

Corollary 3 If in addition to the hypotheses of the theoremwe assume that

int

infin

0

120594119896(119904) 119889119904 le int

infin

120596119896minus1

119889119911

ℎ119896(119911)

119896 = 1 119899 120594119896(119904) = 120582

119896(119904)

120582119896(119904)

(28)

then we have global existence of solutions

Corollary 4 If in addition to the hypotheses of the theoremwe assume that120596

119899(119905) (119899(119905)) grows up at themost polynomially

(or just slower than exp[int1199050119886(119904)119889119904]) then solutions decay at an

exponential rate if int1199050119886(119904)119889119904 rarr infin as 119905 rarr infin

Corollary 5 In addition to the hypotheses of the theoremassume that 1198971015840

119894119895(119905) le 119871

119894119895119897119894119895(119905) 119894 119895 = 1 119898 for some positive

Advances in Artificial Neural Systems 5

constants 119871119894119895and 120595

119894119895(119905) are in the class H (that is 120595

119894119895(120572119906) le

120585119894119895120595119894119895(119906) 120572 gt 0 119906 gt 0 119894 119895 = 1 119898) Then solutions are

bounded by a function of the form exp[minus(int1199050119886(119904)119889119904minus119871119905)] where

119897 = max119871119894119895 119894 119895 = 1 119898

Remark 6 We have assumed that 120572119894119895and 120573

119894119895are greater than

one but the case when they are smaller than one may betreated similarlyWhen their sum is smaller than one we haveglobal existence without adding any extra condition

Remark 7 The decay rate obtained in Corollary 5 is to becompared with the one in the theorem (case (b)) It appearsthat the estimation in Corollary 5 holds for more generalinitial data (not as small as the ones in case (b)) Howeverthe decay rate is smaller than the one in (b) besides assumingthat int1199050119886(119904)119889119904 minus 119871119905 rarr infin as 119905 rarr infin

Remark 8 If we consider the following new functions thenthe monotonicity condition and the order imposed in thetheorem may be dropped

1206011(119905) = max

0le119904le119905

1198921(119904)

120601119896(119905) = max

0le119904le119905

119892119896(119904)

120601119896minus1

(119904)

120601119896minus1

(119905)

(29)

and 120595(119905) = 120601119896(119905)120601119896minus1

(119905)

4 Application

(Artificial)Neural networks are built in an attempt to performdifferent tasks just as the nervous system Typically a neuralnetwork consists of several layers (input layer hidden layersand output layer) Each layer contains one or more cells(neurons) with many connections between them The cellsin one layer receive inputs from the previous layer makesome transformations and send the results to the cells of thesubsequent layer

One may encounter neural networks in many fields suchas control patternmatching settlement of structures classifi-cation of soil supply chain management engineering designmarket segmentation product analysis market developmentforecasting signature verification bond rating recognitionof diseases robust pattern detection text mining price fore-cast botanical classification and scheduling optimization

Neural networks not only can perform many of the tasksa traditional computer can do but also excel in for instanceclassifying incomplete or noisy data predicting future eventsand generalizing

The system (1) is a general version of simpler systems thatappear in neural network theory [1ndash9] like

1199091015840

119894(119905) = minus119886

119894119909119894(119905) +

119898

sum

119895=1

119891119894119895(119909119895(119905)) + 119888

119894(119905) (30)

or

1199091015840

119894(119905) = minus119886

119894119909119894(119905)

+

119898

sum

119895=1

int

119905

minusinfin

119897119894119895(119905 minus 119904) 119891

119894119895(119909119895(119904)) 119889119904 + 119888

119894(119905)

(31)

It is well established by now that (for constant coefficientsand constant 119888

119894(119905)) solutions converge in an exponential

manner to the equilibrium Notice that zero in our case isnot an equilibrium This equilibrium exists and is unique incase of Lipschitz continuity of the activation functions Inour case the system is much more general and the activationfunctions as well as the nonlinearities are not necessarily Lip-schitz continuous However in case of Lipschitz continuityand existence of a unique equilibrium we expect to haveexponential stability using the standard techniques at leastwhen we start away from zero

For the system

1199091015840

119894(119905) = minus119886

119894119909119894(119905)

+

119898

sum

119895=1

119887119894119895

10038161003816100381610038161003816119909119895(119905)

10038161003816100381610038161003816

120572119894119895

(int

119905

minusinfin

119897119894119895(119905 minus 119904) 120595

119894119895(

10038161003816100381610038161003816119909119895(119904)

10038161003816100381610038161003816) 119889119904)

120573119894119895

+ 119888119894(119905)

(32)

(where 120595119894119895

may be taken as power functions see alsoCorollary 5) our theorem gives sufficient conditions guaran-teeing the estimation

119909 (119905) le 120596119899(119905) exp [minusint

119905

0

119886 (119904) 119889119904] 0 le 119905 lt 1205730 (33)

Then Corollaries 3 and 4 provide practical situations wherewe have global existence and decay to zero at an exponentialrate

Conflict of Interests

The author declares that there is no conflict of interestsregarding the publication of this paper

Acknowledgment

The author is grateful for the financial support and thefacilities provided by King Fahd University of Petroleum andMinerals through Grant no IN111052

References

[1] J Cao K Yuan and H-X Li ldquoGlobal asymptotical stabilityof recurrent neural networks with multiple discrete delays anddistributed delaysrdquo IEEE Transactions on Neural Networks vol17 no 6 pp 1646ndash1651 2006

[2] B Crespi ldquoStorage capacity of non-monotonic neuronsrdquoNeuralNetworks vol 12 no 10 pp 1377ndash1389 1999

6 Advances in Artificial Neural Systems

[3] G de Sandre M Forti P Nistri and A Premoli ldquoDynamicalanalysis of full-range cellular neural networks by exploiting dif-ferential variational inequalitiesrdquo IEEE Transactions on Circuitsand Systems I Regular Papers vol 54 no 8 pp 1736ndash1749 2007

[4] C Feng and R Plamondon ldquoOn the stability analysis of delayedneural networks systemsrdquo Neural Networks vol 14 no 9 pp1181ndash1188 2001

[5] J J Hopfield ldquoNeural networks and physical systems withemergent collective computational abilitiesrdquo Proceedings of theNational Academy of Sciences of the United States of Americavol 79 no 8 pp 2554ndash2558 1982

[6] J J Hopfield and D W Tank ldquoComputing with neural circuitsa modelrdquo Science vol 233 no 4764 pp 625ndash633 1986

[7] J I Inoue ldquoRetrieval phase diagrams of non-monotonic Hop-field networksrdquo Journal of Physics A Mathematical and Generalvol 29 no 16 pp 4815ndash4826 1996

[8] B Kosko Neural Network and Fuzzy SystemmdashA DynamicalSystem Approach toMachine Intelligence Prentice-Hall of IndiaNew Delhi India 1991

[9] H-F Yanai and S-I Amari ldquoAuto-associative memory withtwo-stage dynamics of nonmonotonic neuronsrdquo IEEE Transac-tions on Neural Networks vol 7 no 4 pp 803ndash815 1996

[10] X Liu and N Jiang ldquoRobust stability analysis of generalizedneural networks with multiple discrete delays and multipledistributed delaysrdquo Neurocomputing vol 72 no 7ndash9 pp 1789ndash1796 2009

[11] SMohamadKGopalsamy andHAkca ldquoExponential stabilityof artificial neural networks with distributed delays and largeimpulsesrdquo Nonlinear Analysis Real World Applications vol 9no 3 pp 872ndash888 2008

[12] J Park ldquoOn global stability criterion for neural networks withdiscrete and distributed delaysrdquo Chaos Solitons and Fractalsvol 30 no 4 pp 897ndash902 2006

[13] J Park ldquoOn global stability criterion of neural networks withcontinuously distributed delaysrdquo Chaos Solitons and Fractalsvol 37 no 2 pp 444ndash449 2008

[14] Z Qiang M A Run-Nian and X Jin ldquoGlobal exponentialconvergence analysis of Hopfield neural networks with con-tinuously distributed delaysrdquo Communications in TheoreticalPhysics vol 39 no 3 pp 381ndash384 2003

[15] Y Wang W Xiong Q Zhou B Xiao and Y Yu ldquoGlobal expo-nential stability of cellular neural networks with continuouslydistributed delays and impulsesrdquo Physics Letters A vol 350 no1-2 pp 89ndash95 2006

[16] H Wu ldquoGlobal exponential stability of Hopfield neural net-works with delays and inverse Lipschitz neuron activationsrdquoNonlinear Analysis Real World Applications vol 10 no 4 pp2297ndash2306 2009

[17] Q Zhang X P Wei and J Xu ldquoGlobal exponential stability ofHopfield neural networkswith continuously distributed delaysrdquoPhysics Letters A vol 315 no 6 pp 431ndash436 2003

[18] H Zhao ldquoGlobal asymptotic stability of Hopfield neural net-work involving distributed delaysrdquo Neural Networks vol 17 no1 pp 47ndash53 2004

[19] J Zhou S Li and Z Yang ldquoGlobal exponential stability ofHopfield neural networks with distributed delaysrdquo AppliedMathematical Modelling Simulation and Computation for Engi-neering and Environmental Systems vol 33 no 3 pp 1513ndash15202009

[20] R Gavalda and H T Siegelmann ldquoDiscontinuities in recurrentneural networksrdquo Neural Computation vol 11 no 3 pp 715ndash745 1999

[21] H Wu F Tao L Qin R Shi and L He ldquoRobust exponentialstability for interval neural networks with delays and non-Lipschitz activation functionsrdquoNonlinearDynamics vol 66 no4 pp 479ndash487 2011

[22] H Wu and X Xue ldquoStability analysis for neural networks withinverse Lipschitzian neuron activations and impulsesrdquo AppliedMathematical Modelling vol 32 no 11 pp 2347ndash2359 2008

[23] M Forti M Grazzini P Nistri and L Pancioni ldquoGeneralizedLyapunov approach for convergence of neural networks withdiscontinuous or non-LIPschitz activationsrdquo Physica D Nonlin-ear Phenomena vol 214 no 1 pp 88ndash99 2006

[24] N-E Tatar ldquoHopfield neural networks with unbounded mono-tone activation functionsrdquoAdvances inArtificial Neural Systemsvol 2012 Article ID 571358 5 pages 2012

[25] N-E Tatar ldquoControl of systems with Holder continuousfunctions in the distributed delaysrdquo Carpathian Journal ofMathematics vol 30 no 1 pp 123ndash128 2014

[26] G Bao and Z Zeng ldquoAnalysis and design of associative mem-ories based on recurrent neural network with discontinuousactivation functionsrdquoNeurocomputing vol 77 no 1 pp 101ndash1072012

[27] M Forti and P Nistri ldquoGlobal convergence of neural networkswith discontinuous neuron activationsrdquo IEEE Transactions onCircuits and Systems I Fundamental Theory and Applicationsvol 50 no 11 pp 1421ndash1435 2003

[28] YHuang H Zhang and ZWang ldquoDynamical stability analysisof multiple equilibrium points in time-varying delayed recur-rent neural networks with discontinuous activation functionsrdquoNeurocomputing vol 91 pp 21ndash28 2012

[29] L Li and L Huang ldquoDynamical behaviors of a class of recur-rent neural networks with discontinuous neuron activationsrdquoApplied Mathematical Modelling vol 33 no 12 pp 4326ndash43362009

[30] L Li and L Huang ldquoGlobal asymptotic stability of delayedneural networks with discontinuous neuron activationsrdquo Neu-rocomputing vol 72 no 16-18 pp 3726ndash3733 2009

[31] Y Li and H Wu ldquoGlobal stability analysis for periodic solutionin discontinuous neural networks with nonlinear growth acti-vationsrdquo Advances in Difference Equations vol 2009 Article ID798685 14 pages 2009

[32] X Liu and J Cao ldquoRobust state estimation for neural networkswith discontinuous activationsrdquo IEEE Transactions on SystemsMan and Cybernetics B Cybernetics vol 40 no 6 pp 1425ndash1437 2010

[33] J Liu X Liu and W-C Xie ldquoGlobal convergence of neuralnetworks with mixed time-varying delays and discontinuousneuron activationsrdquo Information Sciences vol 183 pp 92ndash1052012

[34] S Qin and X Xue ldquoGlobal exponential stability and global con-vergence in finite time of neural networks with discontinuousactivationsrdquoNeural Processing Letters vol 29 no 3 pp 189ndash2042009

[35] J Wang L Huang and Z Guo ldquoGlobal asymptotic stabilityof neural networks with discontinuous activationsrdquo NeuralNetworks vol 22 no 7 pp 931ndash937 2009

[36] Z Wang L Huang Y Zuo and L Zhang ldquoGlobal robuststability of time-delay systems with discontinuous activationfunctions under polytopic parameter uncertaintiesrdquo Bulletin ofthe KoreanMathematical Society vol 47 no 1 pp 89ndash102 2010

[37] H Wu ldquoGlobal stability analysis of a general class of discontin-uous neural networks with linear growth activation functionsrdquoInformation Sciences vol 179 no 19 pp 3432ndash3441 2009

Advances in Artificial Neural Systems 7

[38] Z Cai and L Huang ldquoExistence and global asymptotic stabilityof periodic solution for discrete and distributed time-varyingdelayed neural networks with discontinuous activationsrdquo Neu-rocomputing vol 74 no 17 pp 3170ndash3179 2011

[39] D Papini and V Taddei ldquoGlobal exponential stability of theperiodic solution of a delayed neural network with discontinu-ous activationsrdquo Physics Letters A vol 343 no 1ndash3 pp 117ndash1282005

[40] M Pinto ldquoIntegral inequalities of Bihari-type and applicationsrdquoFunkcialaj Ekvacioj vol 33 no 3 pp 387ndash403 1990

[41] V Lakshmikhantam and S Leela Differential and IntegralInequalities Theory and Applications vol 55-I of Mathematicsin Sciences and Engineering Edited by Bellman R AcademicPress New York NY USA 1969

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Distributed Sensor Networks

International Journal of

Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014

International Journal of

ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ArtificialNeural Systems

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Computational Intelligence and Neuroscience

Industrial EngineeringJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Page 4: Research Article Long Time Behavior for a System …downloads.hindawi.com/archive/2014/252674.pdf[] J. Cao, K. Yuan, and H.-X. Li, Global asymptotical stability of recurrent neural

4 Advances in Artificial Neural Systems

Therefore119906 (119905) le 119906 (0) + 119888 (119905)

+

119898

sum

119894119895=1

int

119905

0

[119887119894119895(119904) (119906 (119904))

120572119894119895+120573119894119895+ 119897119894119895(0) 120595119894119895(119906 (119904))] 119889119904

119905 gt 0

(18)

where 119906(0) = 119909(0) + sum119898

119894119895=1int

0

minusinfin119897119894119895(minus120590)120595

119894119895(1199090(120590))119889120590 Now we

can apply Lemma 1 to obtain119909 (119905) le 119906 (119905) le 120596

119899(119905) 0 le 119905 lt 120573

0(19)

with 1205960(119905) = 119906(0) + 119888(119905) and 120596

119899(119905) is as in the ldquoPreliminariesrdquo

section(b) Consider 1198971015840

119894119895(119905) 119894 119895 = 1 119898 of arbitrary signs

From expressions (14) and (16) we derive that

119863+119906 (119905) le 119863

+119888 (119905)

+

119898

sum

119894119895=1

[119887119894119895(119905) (119906(119905))

120572119894119895+120573119894119895+ 119897119894119895(0) 120595119894119895(119906 (119905))]

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(120590)

10038161003816100381610038161003816120595119894119895(119906 (119905 minus 120590)) 119889120590 119905 gt 0

(20)

The derivative of the auxiliary function

(119905) = 119906 (119905) +

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816int

119905

119905minus119904

120595119894119895(119906 (120590)) 119889120590 119889119904

119905 ge 0

(21)

is equal to (with the help of (20) and (21))

119863+ (119905) = 119863

+119906 (119905)

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816[120595119894119895(119906 (119905)) minus 120595

119894119895(119906 (119905 minus 119904))] 119889120590 119889119904

le 119863+119888 (119905) +

119898

sum

119894119895=1

[119887119894119895(119905) (119906 (119905))

120572119894119895+120573119894119895+ 119897119894119895(0) 120595119894119895(119906 (119905))]

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(120590)

10038161003816100381610038161003816120595119894119895(119906 (119905 minus 120590)) 119889120590

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816[120595119894119895(119906 (119905)) minus 120595

119894119895(119906 (119905 minus 119904))] 119889119904

le 119863+119888 (119905)

+

119898

sum

119894119895=1

119887119894119895(119905) ( (119905))

120572119894119895+120573119894119895

+[119897119894119895(0) + int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816119889119904]120595119894119895( (119905))

119905 gt 0

(22)

Therefore

(119905) le (0) + 119888 (119905)

+

119898

sum

119894119895=1

int

119905

0

119887119894119895(119904) ( (119904))

120572119894119895+120573119894119895

+[119897119894119895(0) + int

infin

0

100381610038161003816100381610038161198971015840

119894119895(120590)

10038161003816100381610038161003816119889120590]120595

119894119895( (119904)) 119889119904

(23)

with

(0) = 119909 (0) +

119898

sum

119894119895=1

int

0

minusinfin

119897119894119895(minus120590) 120595

119894119895(1199090(120590)) 119889120590

+

119898

sum

119894119895=1

int

infin

0

100381610038161003816100381610038161198971015840

119894119895(119904)

10038161003816100381610038161003816int

0

minus119904

120595119894119895(1199060(120590)) 119889120590 119889119904

1199060(120590) = 119911 (120590) =

119898

sum

119894119895=1

119911119894119895(120590)

=

119898

sum

119894119895=1

int

120590

minusinfin

119897119894119895(120590 minus 120591) 120595

119894119895(1199090(120591)) 119889120591 120590 lt 0

(24)

Applying Lemma 1 to (23) we obtain

119909 (119905) le (119905) le 119899(119905) 0 le 119905 lt 120573

1(25)

and hence

119909 (119905) le 119899(119905) 0 le 119905 lt 120573

1 (26)

where 0(119905) = (0) and

119895(119905) = 119867

minus1

119895[119867119895(119895minus1

(119905)) + int

119905

0

120582119895(119904) 119889119904]

119895 = 1 119899

(27)

and 1205730is chosen so that the functions

119895(119905) 119895 = 1 119899 are

defined for 0 le 119905 lt 1205731

Corollary 3 If in addition to the hypotheses of the theoremwe assume that

int

infin

0

120594119896(119904) 119889119904 le int

infin

120596119896minus1

119889119911

ℎ119896(119911)

119896 = 1 119899 120594119896(119904) = 120582

119896(119904)

120582119896(119904)

(28)

then we have global existence of solutions

Corollary 4 If in addition to the hypotheses of the theoremwe assume that120596

119899(119905) (119899(119905)) grows up at themost polynomially

(or just slower than exp[int1199050119886(119904)119889119904]) then solutions decay at an

exponential rate if int1199050119886(119904)119889119904 rarr infin as 119905 rarr infin

Corollary 5 In addition to the hypotheses of the theoremassume that 1198971015840

119894119895(119905) le 119871

119894119895119897119894119895(119905) 119894 119895 = 1 119898 for some positive

Advances in Artificial Neural Systems 5

constants 119871119894119895and 120595

119894119895(119905) are in the class H (that is 120595

119894119895(120572119906) le

120585119894119895120595119894119895(119906) 120572 gt 0 119906 gt 0 119894 119895 = 1 119898) Then solutions are

bounded by a function of the form exp[minus(int1199050119886(119904)119889119904minus119871119905)] where

119897 = max119871119894119895 119894 119895 = 1 119898

Remark 6 We have assumed that 120572119894119895and 120573

119894119895are greater than

one but the case when they are smaller than one may betreated similarlyWhen their sum is smaller than one we haveglobal existence without adding any extra condition

Remark 7 The decay rate obtained in Corollary 5 is to becompared with the one in the theorem (case (b)) It appearsthat the estimation in Corollary 5 holds for more generalinitial data (not as small as the ones in case (b)) Howeverthe decay rate is smaller than the one in (b) besides assumingthat int1199050119886(119904)119889119904 minus 119871119905 rarr infin as 119905 rarr infin

Remark 8 If we consider the following new functions thenthe monotonicity condition and the order imposed in thetheorem may be dropped

1206011(119905) = max

0le119904le119905

1198921(119904)

120601119896(119905) = max

0le119904le119905

119892119896(119904)

120601119896minus1

(119904)

120601119896minus1

(119905)

(29)

and 120595(119905) = 120601119896(119905)120601119896minus1

(119905)

4 Application

(Artificial)Neural networks are built in an attempt to performdifferent tasks just as the nervous system Typically a neuralnetwork consists of several layers (input layer hidden layersand output layer) Each layer contains one or more cells(neurons) with many connections between them The cellsin one layer receive inputs from the previous layer makesome transformations and send the results to the cells of thesubsequent layer

One may encounter neural networks in many fields suchas control patternmatching settlement of structures classifi-cation of soil supply chain management engineering designmarket segmentation product analysis market developmentforecasting signature verification bond rating recognitionof diseases robust pattern detection text mining price fore-cast botanical classification and scheduling optimization

Neural networks not only can perform many of the tasksa traditional computer can do but also excel in for instanceclassifying incomplete or noisy data predicting future eventsand generalizing

The system (1) is a general version of simpler systems thatappear in neural network theory [1ndash9] like

1199091015840

119894(119905) = minus119886

119894119909119894(119905) +

119898

sum

119895=1

119891119894119895(119909119895(119905)) + 119888

119894(119905) (30)

or

1199091015840

119894(119905) = minus119886

119894119909119894(119905)

+

119898

sum

119895=1

int

119905

minusinfin

119897119894119895(119905 minus 119904) 119891

119894119895(119909119895(119904)) 119889119904 + 119888

119894(119905)

(31)

It is well established by now that (for constant coefficientsand constant 119888

119894(119905)) solutions converge in an exponential

manner to the equilibrium Notice that zero in our case isnot an equilibrium This equilibrium exists and is unique incase of Lipschitz continuity of the activation functions Inour case the system is much more general and the activationfunctions as well as the nonlinearities are not necessarily Lip-schitz continuous However in case of Lipschitz continuityand existence of a unique equilibrium we expect to haveexponential stability using the standard techniques at leastwhen we start away from zero

For the system

1199091015840

119894(119905) = minus119886

119894119909119894(119905)

+

119898

sum

119895=1

119887119894119895

10038161003816100381610038161003816119909119895(119905)

10038161003816100381610038161003816

120572119894119895

(int

119905

minusinfin

119897119894119895(119905 minus 119904) 120595

119894119895(

10038161003816100381610038161003816119909119895(119904)

10038161003816100381610038161003816) 119889119904)

120573119894119895

+ 119888119894(119905)

(32)

(where 120595119894119895

may be taken as power functions see alsoCorollary 5) our theorem gives sufficient conditions guaran-teeing the estimation

119909 (119905) le 120596119899(119905) exp [minusint

119905

0

119886 (119904) 119889119904] 0 le 119905 lt 1205730 (33)

Then Corollaries 3 and 4 provide practical situations wherewe have global existence and decay to zero at an exponentialrate

Conflict of Interests

The author declares that there is no conflict of interestsregarding the publication of this paper

Acknowledgment

The author is grateful for the financial support and thefacilities provided by King Fahd University of Petroleum andMinerals through Grant no IN111052

References

[1] J Cao K Yuan and H-X Li ldquoGlobal asymptotical stabilityof recurrent neural networks with multiple discrete delays anddistributed delaysrdquo IEEE Transactions on Neural Networks vol17 no 6 pp 1646ndash1651 2006

[2] B Crespi ldquoStorage capacity of non-monotonic neuronsrdquoNeuralNetworks vol 12 no 10 pp 1377ndash1389 1999

6 Advances in Artificial Neural Systems

[3] G de Sandre M Forti P Nistri and A Premoli ldquoDynamicalanalysis of full-range cellular neural networks by exploiting dif-ferential variational inequalitiesrdquo IEEE Transactions on Circuitsand Systems I Regular Papers vol 54 no 8 pp 1736ndash1749 2007

[4] C Feng and R Plamondon ldquoOn the stability analysis of delayedneural networks systemsrdquo Neural Networks vol 14 no 9 pp1181ndash1188 2001

[5] J J Hopfield ldquoNeural networks and physical systems withemergent collective computational abilitiesrdquo Proceedings of theNational Academy of Sciences of the United States of Americavol 79 no 8 pp 2554ndash2558 1982

[6] J J Hopfield and D W Tank ldquoComputing with neural circuitsa modelrdquo Science vol 233 no 4764 pp 625ndash633 1986

[7] J I Inoue ldquoRetrieval phase diagrams of non-monotonic Hop-field networksrdquo Journal of Physics A Mathematical and Generalvol 29 no 16 pp 4815ndash4826 1996

[8] B Kosko Neural Network and Fuzzy SystemmdashA DynamicalSystem Approach toMachine Intelligence Prentice-Hall of IndiaNew Delhi India 1991

[9] H-F Yanai and S-I Amari ldquoAuto-associative memory withtwo-stage dynamics of nonmonotonic neuronsrdquo IEEE Transac-tions on Neural Networks vol 7 no 4 pp 803ndash815 1996

[10] X Liu and N Jiang ldquoRobust stability analysis of generalizedneural networks with multiple discrete delays and multipledistributed delaysrdquo Neurocomputing vol 72 no 7ndash9 pp 1789ndash1796 2009

[11] SMohamadKGopalsamy andHAkca ldquoExponential stabilityof artificial neural networks with distributed delays and largeimpulsesrdquo Nonlinear Analysis Real World Applications vol 9no 3 pp 872ndash888 2008

[12] J Park ldquoOn global stability criterion for neural networks withdiscrete and distributed delaysrdquo Chaos Solitons and Fractalsvol 30 no 4 pp 897ndash902 2006

[13] J Park ldquoOn global stability criterion of neural networks withcontinuously distributed delaysrdquo Chaos Solitons and Fractalsvol 37 no 2 pp 444ndash449 2008

[14] Z Qiang M A Run-Nian and X Jin ldquoGlobal exponentialconvergence analysis of Hopfield neural networks with con-tinuously distributed delaysrdquo Communications in TheoreticalPhysics vol 39 no 3 pp 381ndash384 2003

[15] Y Wang W Xiong Q Zhou B Xiao and Y Yu ldquoGlobal expo-nential stability of cellular neural networks with continuouslydistributed delays and impulsesrdquo Physics Letters A vol 350 no1-2 pp 89ndash95 2006

[16] H Wu ldquoGlobal exponential stability of Hopfield neural net-works with delays and inverse Lipschitz neuron activationsrdquoNonlinear Analysis Real World Applications vol 10 no 4 pp2297ndash2306 2009

[17] Q Zhang X P Wei and J Xu ldquoGlobal exponential stability ofHopfield neural networkswith continuously distributed delaysrdquoPhysics Letters A vol 315 no 6 pp 431ndash436 2003

[18] H Zhao ldquoGlobal asymptotic stability of Hopfield neural net-work involving distributed delaysrdquo Neural Networks vol 17 no1 pp 47ndash53 2004

[19] J Zhou S Li and Z Yang ldquoGlobal exponential stability ofHopfield neural networks with distributed delaysrdquo AppliedMathematical Modelling Simulation and Computation for Engi-neering and Environmental Systems vol 33 no 3 pp 1513ndash15202009

[20] R Gavalda and H T Siegelmann ldquoDiscontinuities in recurrentneural networksrdquo Neural Computation vol 11 no 3 pp 715ndash745 1999

[21] H Wu F Tao L Qin R Shi and L He ldquoRobust exponentialstability for interval neural networks with delays and non-Lipschitz activation functionsrdquoNonlinearDynamics vol 66 no4 pp 479ndash487 2011

[22] H Wu and X Xue ldquoStability analysis for neural networks withinverse Lipschitzian neuron activations and impulsesrdquo AppliedMathematical Modelling vol 32 no 11 pp 2347ndash2359 2008

[23] M Forti M Grazzini P Nistri and L Pancioni ldquoGeneralizedLyapunov approach for convergence of neural networks withdiscontinuous or non-LIPschitz activationsrdquo Physica D Nonlin-ear Phenomena vol 214 no 1 pp 88ndash99 2006

[24] N-E Tatar ldquoHopfield neural networks with unbounded mono-tone activation functionsrdquoAdvances inArtificial Neural Systemsvol 2012 Article ID 571358 5 pages 2012

[25] N-E Tatar ldquoControl of systems with Holder continuousfunctions in the distributed delaysrdquo Carpathian Journal ofMathematics vol 30 no 1 pp 123ndash128 2014

[26] G Bao and Z Zeng ldquoAnalysis and design of associative mem-ories based on recurrent neural network with discontinuousactivation functionsrdquoNeurocomputing vol 77 no 1 pp 101ndash1072012

[27] M Forti and P Nistri ldquoGlobal convergence of neural networkswith discontinuous neuron activationsrdquo IEEE Transactions onCircuits and Systems I Fundamental Theory and Applicationsvol 50 no 11 pp 1421ndash1435 2003

[28] YHuang H Zhang and ZWang ldquoDynamical stability analysisof multiple equilibrium points in time-varying delayed recur-rent neural networks with discontinuous activation functionsrdquoNeurocomputing vol 91 pp 21ndash28 2012

[29] L Li and L Huang ldquoDynamical behaviors of a class of recur-rent neural networks with discontinuous neuron activationsrdquoApplied Mathematical Modelling vol 33 no 12 pp 4326ndash43362009

[30] L Li and L Huang ldquoGlobal asymptotic stability of delayedneural networks with discontinuous neuron activationsrdquo Neu-rocomputing vol 72 no 16-18 pp 3726ndash3733 2009

[31] Y Li and H Wu ldquoGlobal stability analysis for periodic solutionin discontinuous neural networks with nonlinear growth acti-vationsrdquo Advances in Difference Equations vol 2009 Article ID798685 14 pages 2009

[32] X Liu and J Cao ldquoRobust state estimation for neural networkswith discontinuous activationsrdquo IEEE Transactions on SystemsMan and Cybernetics B Cybernetics vol 40 no 6 pp 1425ndash1437 2010

[33] J Liu X Liu and W-C Xie ldquoGlobal convergence of neuralnetworks with mixed time-varying delays and discontinuousneuron activationsrdquo Information Sciences vol 183 pp 92ndash1052012

[34] S Qin and X Xue ldquoGlobal exponential stability and global con-vergence in finite time of neural networks with discontinuousactivationsrdquoNeural Processing Letters vol 29 no 3 pp 189ndash2042009

[35] J Wang L Huang and Z Guo ldquoGlobal asymptotic stabilityof neural networks with discontinuous activationsrdquo NeuralNetworks vol 22 no 7 pp 931ndash937 2009

[36] Z Wang L Huang Y Zuo and L Zhang ldquoGlobal robuststability of time-delay systems with discontinuous activationfunctions under polytopic parameter uncertaintiesrdquo Bulletin ofthe KoreanMathematical Society vol 47 no 1 pp 89ndash102 2010

[37] H Wu ldquoGlobal stability analysis of a general class of discontin-uous neural networks with linear growth activation functionsrdquoInformation Sciences vol 179 no 19 pp 3432ndash3441 2009

Advances in Artificial Neural Systems 7

[38] Z Cai and L Huang ldquoExistence and global asymptotic stabilityof periodic solution for discrete and distributed time-varyingdelayed neural networks with discontinuous activationsrdquo Neu-rocomputing vol 74 no 17 pp 3170ndash3179 2011

[39] D Papini and V Taddei ldquoGlobal exponential stability of theperiodic solution of a delayed neural network with discontinu-ous activationsrdquo Physics Letters A vol 343 no 1ndash3 pp 117ndash1282005

[40] M Pinto ldquoIntegral inequalities of Bihari-type and applicationsrdquoFunkcialaj Ekvacioj vol 33 no 3 pp 387ndash403 1990

[41] V Lakshmikhantam and S Leela Differential and IntegralInequalities Theory and Applications vol 55-I of Mathematicsin Sciences and Engineering Edited by Bellman R AcademicPress New York NY USA 1969

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Distributed Sensor Networks

International Journal of

Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014

International Journal of

ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ArtificialNeural Systems

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Computational Intelligence and Neuroscience

Industrial EngineeringJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Page 5: Research Article Long Time Behavior for a System …downloads.hindawi.com/archive/2014/252674.pdf[] J. Cao, K. Yuan, and H.-X. Li, Global asymptotical stability of recurrent neural

Advances in Artificial Neural Systems 5

constants 119871119894119895and 120595

119894119895(119905) are in the class H (that is 120595

119894119895(120572119906) le

120585119894119895120595119894119895(119906) 120572 gt 0 119906 gt 0 119894 119895 = 1 119898) Then solutions are

bounded by a function of the form exp[minus(int1199050119886(119904)119889119904minus119871119905)] where

119897 = max119871119894119895 119894 119895 = 1 119898

Remark 6 We have assumed that 120572119894119895and 120573

119894119895are greater than

one but the case when they are smaller than one may betreated similarlyWhen their sum is smaller than one we haveglobal existence without adding any extra condition

Remark 7 The decay rate obtained in Corollary 5 is to becompared with the one in the theorem (case (b)) It appearsthat the estimation in Corollary 5 holds for more generalinitial data (not as small as the ones in case (b)) Howeverthe decay rate is smaller than the one in (b) besides assumingthat int1199050119886(119904)119889119904 minus 119871119905 rarr infin as 119905 rarr infin

Remark 8 If we consider the following new functions thenthe monotonicity condition and the order imposed in thetheorem may be dropped

1206011(119905) = max

0le119904le119905

1198921(119904)

120601119896(119905) = max

0le119904le119905

119892119896(119904)

120601119896minus1

(119904)

120601119896minus1

(119905)

(29)

and 120595(119905) = 120601119896(119905)120601119896minus1

(119905)

4 Application

(Artificial)Neural networks are built in an attempt to performdifferent tasks just as the nervous system Typically a neuralnetwork consists of several layers (input layer hidden layersand output layer) Each layer contains one or more cells(neurons) with many connections between them The cellsin one layer receive inputs from the previous layer makesome transformations and send the results to the cells of thesubsequent layer

One may encounter neural networks in many fields suchas control patternmatching settlement of structures classifi-cation of soil supply chain management engineering designmarket segmentation product analysis market developmentforecasting signature verification bond rating recognitionof diseases robust pattern detection text mining price fore-cast botanical classification and scheduling optimization

Neural networks not only can perform many of the tasksa traditional computer can do but also excel in for instanceclassifying incomplete or noisy data predicting future eventsand generalizing

The system (1) is a general version of simpler systems thatappear in neural network theory [1ndash9] like

1199091015840

119894(119905) = minus119886

119894119909119894(119905) +

119898

sum

119895=1

119891119894119895(119909119895(119905)) + 119888

119894(119905) (30)

or

1199091015840

119894(119905) = minus119886

119894119909119894(119905)

+

119898

sum

119895=1

int

119905

minusinfin

119897119894119895(119905 minus 119904) 119891

119894119895(119909119895(119904)) 119889119904 + 119888

119894(119905)

(31)

It is well established by now that (for constant coefficientsand constant 119888

119894(119905)) solutions converge in an exponential

manner to the equilibrium Notice that zero in our case isnot an equilibrium This equilibrium exists and is unique incase of Lipschitz continuity of the activation functions Inour case the system is much more general and the activationfunctions as well as the nonlinearities are not necessarily Lip-schitz continuous However in case of Lipschitz continuityand existence of a unique equilibrium we expect to haveexponential stability using the standard techniques at leastwhen we start away from zero

For the system

1199091015840

119894(119905) = minus119886

119894119909119894(119905)

+

119898

sum

119895=1

119887119894119895

10038161003816100381610038161003816119909119895(119905)

10038161003816100381610038161003816

120572119894119895

(int

119905

minusinfin

119897119894119895(119905 minus 119904) 120595

119894119895(

10038161003816100381610038161003816119909119895(119904)

10038161003816100381610038161003816) 119889119904)

120573119894119895

+ 119888119894(119905)

(32)

(where 120595119894119895

may be taken as power functions see alsoCorollary 5) our theorem gives sufficient conditions guaran-teeing the estimation

119909 (119905) le 120596119899(119905) exp [minusint

119905

0

119886 (119904) 119889119904] 0 le 119905 lt 1205730 (33)

Then Corollaries 3 and 4 provide practical situations wherewe have global existence and decay to zero at an exponentialrate

Conflict of Interests

The author declares that there is no conflict of interestsregarding the publication of this paper

Acknowledgment

The author is grateful for the financial support and thefacilities provided by King Fahd University of Petroleum andMinerals through Grant no IN111052

References

[1] J Cao K Yuan and H-X Li ldquoGlobal asymptotical stabilityof recurrent neural networks with multiple discrete delays anddistributed delaysrdquo IEEE Transactions on Neural Networks vol17 no 6 pp 1646ndash1651 2006

[2] B Crespi ldquoStorage capacity of non-monotonic neuronsrdquoNeuralNetworks vol 12 no 10 pp 1377ndash1389 1999

6 Advances in Artificial Neural Systems

[3] G de Sandre M Forti P Nistri and A Premoli ldquoDynamicalanalysis of full-range cellular neural networks by exploiting dif-ferential variational inequalitiesrdquo IEEE Transactions on Circuitsand Systems I Regular Papers vol 54 no 8 pp 1736ndash1749 2007

[4] C Feng and R Plamondon ldquoOn the stability analysis of delayedneural networks systemsrdquo Neural Networks vol 14 no 9 pp1181ndash1188 2001

[5] J J Hopfield ldquoNeural networks and physical systems withemergent collective computational abilitiesrdquo Proceedings of theNational Academy of Sciences of the United States of Americavol 79 no 8 pp 2554ndash2558 1982

[6] J J Hopfield and D W Tank ldquoComputing with neural circuitsa modelrdquo Science vol 233 no 4764 pp 625ndash633 1986

[7] J I Inoue ldquoRetrieval phase diagrams of non-monotonic Hop-field networksrdquo Journal of Physics A Mathematical and Generalvol 29 no 16 pp 4815ndash4826 1996

[8] B Kosko Neural Network and Fuzzy SystemmdashA DynamicalSystem Approach toMachine Intelligence Prentice-Hall of IndiaNew Delhi India 1991

[9] H-F Yanai and S-I Amari ldquoAuto-associative memory withtwo-stage dynamics of nonmonotonic neuronsrdquo IEEE Transac-tions on Neural Networks vol 7 no 4 pp 803ndash815 1996

[10] X Liu and N Jiang ldquoRobust stability analysis of generalizedneural networks with multiple discrete delays and multipledistributed delaysrdquo Neurocomputing vol 72 no 7ndash9 pp 1789ndash1796 2009

[11] SMohamadKGopalsamy andHAkca ldquoExponential stabilityof artificial neural networks with distributed delays and largeimpulsesrdquo Nonlinear Analysis Real World Applications vol 9no 3 pp 872ndash888 2008

[12] J Park ldquoOn global stability criterion for neural networks withdiscrete and distributed delaysrdquo Chaos Solitons and Fractalsvol 30 no 4 pp 897ndash902 2006

[13] J Park ldquoOn global stability criterion of neural networks withcontinuously distributed delaysrdquo Chaos Solitons and Fractalsvol 37 no 2 pp 444ndash449 2008

[14] Z Qiang M A Run-Nian and X Jin ldquoGlobal exponentialconvergence analysis of Hopfield neural networks with con-tinuously distributed delaysrdquo Communications in TheoreticalPhysics vol 39 no 3 pp 381ndash384 2003

[15] Y Wang W Xiong Q Zhou B Xiao and Y Yu ldquoGlobal expo-nential stability of cellular neural networks with continuouslydistributed delays and impulsesrdquo Physics Letters A vol 350 no1-2 pp 89ndash95 2006

[16] H Wu ldquoGlobal exponential stability of Hopfield neural net-works with delays and inverse Lipschitz neuron activationsrdquoNonlinear Analysis Real World Applications vol 10 no 4 pp2297ndash2306 2009

[17] Q Zhang X P Wei and J Xu ldquoGlobal exponential stability ofHopfield neural networkswith continuously distributed delaysrdquoPhysics Letters A vol 315 no 6 pp 431ndash436 2003

[18] H Zhao ldquoGlobal asymptotic stability of Hopfield neural net-work involving distributed delaysrdquo Neural Networks vol 17 no1 pp 47ndash53 2004

[19] J Zhou S Li and Z Yang ldquoGlobal exponential stability ofHopfield neural networks with distributed delaysrdquo AppliedMathematical Modelling Simulation and Computation for Engi-neering and Environmental Systems vol 33 no 3 pp 1513ndash15202009

[20] R Gavalda and H T Siegelmann ldquoDiscontinuities in recurrentneural networksrdquo Neural Computation vol 11 no 3 pp 715ndash745 1999

[21] H Wu F Tao L Qin R Shi and L He ldquoRobust exponentialstability for interval neural networks with delays and non-Lipschitz activation functionsrdquoNonlinearDynamics vol 66 no4 pp 479ndash487 2011

[22] H Wu and X Xue ldquoStability analysis for neural networks withinverse Lipschitzian neuron activations and impulsesrdquo AppliedMathematical Modelling vol 32 no 11 pp 2347ndash2359 2008

[23] M Forti M Grazzini P Nistri and L Pancioni ldquoGeneralizedLyapunov approach for convergence of neural networks withdiscontinuous or non-LIPschitz activationsrdquo Physica D Nonlin-ear Phenomena vol 214 no 1 pp 88ndash99 2006

[24] N-E Tatar ldquoHopfield neural networks with unbounded mono-tone activation functionsrdquoAdvances inArtificial Neural Systemsvol 2012 Article ID 571358 5 pages 2012

[25] N-E Tatar ldquoControl of systems with Holder continuousfunctions in the distributed delaysrdquo Carpathian Journal ofMathematics vol 30 no 1 pp 123ndash128 2014

[26] G Bao and Z Zeng ldquoAnalysis and design of associative mem-ories based on recurrent neural network with discontinuousactivation functionsrdquoNeurocomputing vol 77 no 1 pp 101ndash1072012

[27] M Forti and P Nistri ldquoGlobal convergence of neural networkswith discontinuous neuron activationsrdquo IEEE Transactions onCircuits and Systems I Fundamental Theory and Applicationsvol 50 no 11 pp 1421ndash1435 2003

[28] YHuang H Zhang and ZWang ldquoDynamical stability analysisof multiple equilibrium points in time-varying delayed recur-rent neural networks with discontinuous activation functionsrdquoNeurocomputing vol 91 pp 21ndash28 2012

[29] L Li and L Huang ldquoDynamical behaviors of a class of recur-rent neural networks with discontinuous neuron activationsrdquoApplied Mathematical Modelling vol 33 no 12 pp 4326ndash43362009

[30] L Li and L Huang ldquoGlobal asymptotic stability of delayedneural networks with discontinuous neuron activationsrdquo Neu-rocomputing vol 72 no 16-18 pp 3726ndash3733 2009

[31] Y Li and H Wu ldquoGlobal stability analysis for periodic solutionin discontinuous neural networks with nonlinear growth acti-vationsrdquo Advances in Difference Equations vol 2009 Article ID798685 14 pages 2009

[32] X Liu and J Cao ldquoRobust state estimation for neural networkswith discontinuous activationsrdquo IEEE Transactions on SystemsMan and Cybernetics B Cybernetics vol 40 no 6 pp 1425ndash1437 2010

[33] J Liu X Liu and W-C Xie ldquoGlobal convergence of neuralnetworks with mixed time-varying delays and discontinuousneuron activationsrdquo Information Sciences vol 183 pp 92ndash1052012

[34] S Qin and X Xue ldquoGlobal exponential stability and global con-vergence in finite time of neural networks with discontinuousactivationsrdquoNeural Processing Letters vol 29 no 3 pp 189ndash2042009

[35] J Wang L Huang and Z Guo ldquoGlobal asymptotic stabilityof neural networks with discontinuous activationsrdquo NeuralNetworks vol 22 no 7 pp 931ndash937 2009

[36] Z Wang L Huang Y Zuo and L Zhang ldquoGlobal robuststability of time-delay systems with discontinuous activationfunctions under polytopic parameter uncertaintiesrdquo Bulletin ofthe KoreanMathematical Society vol 47 no 1 pp 89ndash102 2010

[37] H Wu ldquoGlobal stability analysis of a general class of discontin-uous neural networks with linear growth activation functionsrdquoInformation Sciences vol 179 no 19 pp 3432ndash3441 2009

Advances in Artificial Neural Systems 7

[38] Z Cai and L Huang ldquoExistence and global asymptotic stabilityof periodic solution for discrete and distributed time-varyingdelayed neural networks with discontinuous activationsrdquo Neu-rocomputing vol 74 no 17 pp 3170ndash3179 2011

[39] D Papini and V Taddei ldquoGlobal exponential stability of theperiodic solution of a delayed neural network with discontinu-ous activationsrdquo Physics Letters A vol 343 no 1ndash3 pp 117ndash1282005

[40] M Pinto ldquoIntegral inequalities of Bihari-type and applicationsrdquoFunkcialaj Ekvacioj vol 33 no 3 pp 387ndash403 1990

[41] V Lakshmikhantam and S Leela Differential and IntegralInequalities Theory and Applications vol 55-I of Mathematicsin Sciences and Engineering Edited by Bellman R AcademicPress New York NY USA 1969

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Distributed Sensor Networks

International Journal of

Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014

International Journal of

ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ArtificialNeural Systems

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Computational Intelligence and Neuroscience

Industrial EngineeringJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Page 6: Research Article Long Time Behavior for a System …downloads.hindawi.com/archive/2014/252674.pdf[] J. Cao, K. Yuan, and H.-X. Li, Global asymptotical stability of recurrent neural

6 Advances in Artificial Neural Systems

[3] G de Sandre M Forti P Nistri and A Premoli ldquoDynamicalanalysis of full-range cellular neural networks by exploiting dif-ferential variational inequalitiesrdquo IEEE Transactions on Circuitsand Systems I Regular Papers vol 54 no 8 pp 1736ndash1749 2007

[4] C Feng and R Plamondon ldquoOn the stability analysis of delayedneural networks systemsrdquo Neural Networks vol 14 no 9 pp1181ndash1188 2001

[5] J J Hopfield ldquoNeural networks and physical systems withemergent collective computational abilitiesrdquo Proceedings of theNational Academy of Sciences of the United States of Americavol 79 no 8 pp 2554ndash2558 1982

[6] J J Hopfield and D W Tank ldquoComputing with neural circuitsa modelrdquo Science vol 233 no 4764 pp 625ndash633 1986

[7] J I Inoue ldquoRetrieval phase diagrams of non-monotonic Hop-field networksrdquo Journal of Physics A Mathematical and Generalvol 29 no 16 pp 4815ndash4826 1996

[8] B Kosko Neural Network and Fuzzy SystemmdashA DynamicalSystem Approach toMachine Intelligence Prentice-Hall of IndiaNew Delhi India 1991

[9] H-F Yanai and S-I Amari ldquoAuto-associative memory withtwo-stage dynamics of nonmonotonic neuronsrdquo IEEE Transac-tions on Neural Networks vol 7 no 4 pp 803ndash815 1996

[10] X Liu and N Jiang ldquoRobust stability analysis of generalizedneural networks with multiple discrete delays and multipledistributed delaysrdquo Neurocomputing vol 72 no 7ndash9 pp 1789ndash1796 2009

[11] SMohamadKGopalsamy andHAkca ldquoExponential stabilityof artificial neural networks with distributed delays and largeimpulsesrdquo Nonlinear Analysis Real World Applications vol 9no 3 pp 872ndash888 2008

[12] J Park ldquoOn global stability criterion for neural networks withdiscrete and distributed delaysrdquo Chaos Solitons and Fractalsvol 30 no 4 pp 897ndash902 2006

[13] J Park ldquoOn global stability criterion of neural networks withcontinuously distributed delaysrdquo Chaos Solitons and Fractalsvol 37 no 2 pp 444ndash449 2008

[14] Z Qiang M A Run-Nian and X Jin ldquoGlobal exponentialconvergence analysis of Hopfield neural networks with con-tinuously distributed delaysrdquo Communications in TheoreticalPhysics vol 39 no 3 pp 381ndash384 2003

[15] Y Wang W Xiong Q Zhou B Xiao and Y Yu ldquoGlobal expo-nential stability of cellular neural networks with continuouslydistributed delays and impulsesrdquo Physics Letters A vol 350 no1-2 pp 89ndash95 2006

[16] H Wu ldquoGlobal exponential stability of Hopfield neural net-works with delays and inverse Lipschitz neuron activationsrdquoNonlinear Analysis Real World Applications vol 10 no 4 pp2297ndash2306 2009

[17] Q Zhang X P Wei and J Xu ldquoGlobal exponential stability ofHopfield neural networkswith continuously distributed delaysrdquoPhysics Letters A vol 315 no 6 pp 431ndash436 2003

[18] H Zhao ldquoGlobal asymptotic stability of Hopfield neural net-work involving distributed delaysrdquo Neural Networks vol 17 no1 pp 47ndash53 2004

[19] J Zhou S Li and Z Yang ldquoGlobal exponential stability ofHopfield neural networks with distributed delaysrdquo AppliedMathematical Modelling Simulation and Computation for Engi-neering and Environmental Systems vol 33 no 3 pp 1513ndash15202009

[20] R Gavalda and H T Siegelmann ldquoDiscontinuities in recurrentneural networksrdquo Neural Computation vol 11 no 3 pp 715ndash745 1999

[21] H Wu F Tao L Qin R Shi and L He ldquoRobust exponentialstability for interval neural networks with delays and non-Lipschitz activation functionsrdquoNonlinearDynamics vol 66 no4 pp 479ndash487 2011

[22] H Wu and X Xue ldquoStability analysis for neural networks withinverse Lipschitzian neuron activations and impulsesrdquo AppliedMathematical Modelling vol 32 no 11 pp 2347ndash2359 2008

[23] M Forti M Grazzini P Nistri and L Pancioni ldquoGeneralizedLyapunov approach for convergence of neural networks withdiscontinuous or non-LIPschitz activationsrdquo Physica D Nonlin-ear Phenomena vol 214 no 1 pp 88ndash99 2006

[24] N-E Tatar ldquoHopfield neural networks with unbounded mono-tone activation functionsrdquoAdvances inArtificial Neural Systemsvol 2012 Article ID 571358 5 pages 2012

[25] N-E Tatar ldquoControl of systems with Holder continuousfunctions in the distributed delaysrdquo Carpathian Journal ofMathematics vol 30 no 1 pp 123ndash128 2014

[26] G Bao and Z Zeng ldquoAnalysis and design of associative mem-ories based on recurrent neural network with discontinuousactivation functionsrdquoNeurocomputing vol 77 no 1 pp 101ndash1072012

[27] M Forti and P Nistri ldquoGlobal convergence of neural networkswith discontinuous neuron activationsrdquo IEEE Transactions onCircuits and Systems I Fundamental Theory and Applicationsvol 50 no 11 pp 1421ndash1435 2003

[28] YHuang H Zhang and ZWang ldquoDynamical stability analysisof multiple equilibrium points in time-varying delayed recur-rent neural networks with discontinuous activation functionsrdquoNeurocomputing vol 91 pp 21ndash28 2012

[29] L Li and L Huang ldquoDynamical behaviors of a class of recur-rent neural networks with discontinuous neuron activationsrdquoApplied Mathematical Modelling vol 33 no 12 pp 4326ndash43362009

[30] L Li and L Huang ldquoGlobal asymptotic stability of delayedneural networks with discontinuous neuron activationsrdquo Neu-rocomputing vol 72 no 16-18 pp 3726ndash3733 2009

[31] Y Li and H Wu ldquoGlobal stability analysis for periodic solutionin discontinuous neural networks with nonlinear growth acti-vationsrdquo Advances in Difference Equations vol 2009 Article ID798685 14 pages 2009

[32] X Liu and J Cao ldquoRobust state estimation for neural networkswith discontinuous activationsrdquo IEEE Transactions on SystemsMan and Cybernetics B Cybernetics vol 40 no 6 pp 1425ndash1437 2010

[33] J Liu X Liu and W-C Xie ldquoGlobal convergence of neuralnetworks with mixed time-varying delays and discontinuousneuron activationsrdquo Information Sciences vol 183 pp 92ndash1052012

[34] S Qin and X Xue ldquoGlobal exponential stability and global con-vergence in finite time of neural networks with discontinuousactivationsrdquoNeural Processing Letters vol 29 no 3 pp 189ndash2042009

[35] J Wang L Huang and Z Guo ldquoGlobal asymptotic stabilityof neural networks with discontinuous activationsrdquo NeuralNetworks vol 22 no 7 pp 931ndash937 2009

[36] Z Wang L Huang Y Zuo and L Zhang ldquoGlobal robuststability of time-delay systems with discontinuous activationfunctions under polytopic parameter uncertaintiesrdquo Bulletin ofthe KoreanMathematical Society vol 47 no 1 pp 89ndash102 2010

[37] H Wu ldquoGlobal stability analysis of a general class of discontin-uous neural networks with linear growth activation functionsrdquoInformation Sciences vol 179 no 19 pp 3432ndash3441 2009

Advances in Artificial Neural Systems 7

[38] Z Cai and L Huang ldquoExistence and global asymptotic stabilityof periodic solution for discrete and distributed time-varyingdelayed neural networks with discontinuous activationsrdquo Neu-rocomputing vol 74 no 17 pp 3170ndash3179 2011

[39] D Papini and V Taddei ldquoGlobal exponential stability of theperiodic solution of a delayed neural network with discontinu-ous activationsrdquo Physics Letters A vol 343 no 1ndash3 pp 117ndash1282005

[40] M Pinto ldquoIntegral inequalities of Bihari-type and applicationsrdquoFunkcialaj Ekvacioj vol 33 no 3 pp 387ndash403 1990

[41] V Lakshmikhantam and S Leela Differential and IntegralInequalities Theory and Applications vol 55-I of Mathematicsin Sciences and Engineering Edited by Bellman R AcademicPress New York NY USA 1969

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Distributed Sensor Networks

International Journal of

Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014

International Journal of

ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ArtificialNeural Systems

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Computational Intelligence and Neuroscience

Industrial EngineeringJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Page 7: Research Article Long Time Behavior for a System …downloads.hindawi.com/archive/2014/252674.pdf[] J. Cao, K. Yuan, and H.-X. Li, Global asymptotical stability of recurrent neural

Advances in Artificial Neural Systems 7

[38] Z Cai and L Huang ldquoExistence and global asymptotic stabilityof periodic solution for discrete and distributed time-varyingdelayed neural networks with discontinuous activationsrdquo Neu-rocomputing vol 74 no 17 pp 3170ndash3179 2011

[39] D Papini and V Taddei ldquoGlobal exponential stability of theperiodic solution of a delayed neural network with discontinu-ous activationsrdquo Physics Letters A vol 343 no 1ndash3 pp 117ndash1282005

[40] M Pinto ldquoIntegral inequalities of Bihari-type and applicationsrdquoFunkcialaj Ekvacioj vol 33 no 3 pp 387ndash403 1990

[41] V Lakshmikhantam and S Leela Differential and IntegralInequalities Theory and Applications vol 55-I of Mathematicsin Sciences and Engineering Edited by Bellman R AcademicPress New York NY USA 1969

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Distributed Sensor Networks

International Journal of

Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014

International Journal of

ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ArtificialNeural Systems

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Computational Intelligence and Neuroscience

Industrial EngineeringJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Page 8: Research Article Long Time Behavior for a System …downloads.hindawi.com/archive/2014/252674.pdf[] J. Cao, K. Yuan, and H.-X. Li, Global asymptotical stability of recurrent neural

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Distributed Sensor Networks

International Journal of

Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014

International Journal of

ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia

International Journal of

Biomedical Imaging

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

ArtificialNeural Systems

Advances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Computational Intelligence and Neuroscience

Industrial EngineeringJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014

Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Recommended