姚慧麗++孫海彤++卜憲江
摘要:自一類特殊的微分方程即帶有延遲的細(xì)胞神經(jīng)網(wǎng)絡(luò)被建立以來,討論其各種解的存在性和唯一性成為重要研究內(nèi)容,特別是對概周期型解的研究。為了探討漸近概周期函數(shù)在一類帶可變延遲細(xì)胞神經(jīng)網(wǎng)絡(luò)中的應(yīng)用,依據(jù)Banach不動點(diǎn)定理和指數(shù)二分法的有關(guān)知識,并結(jié)合已有文獻(xiàn)對這類細(xì)胞神經(jīng)網(wǎng)絡(luò)概周期解的研究,討論了這類細(xì)胞神經(jīng)網(wǎng)絡(luò)的漸近概周期解的存在及唯一性問題,證明了該類方程在一定的充分條件下有唯一的漸近概周期解。
關(guān)鍵詞:
細(xì)胞神經(jīng)網(wǎng)絡(luò);漸近概周期解;指數(shù)二分法;不動點(diǎn);延遲
DOI:1015938/jjhust201705024
中圖分類號: O175
文獻(xiàn)標(biāo)志碼: A
文章編號: 1007-2683(2017)05-0130-07
Asymptotically Almost Periodic Solutions for a Class
of Cellular Neural Networks with Varying Delays
YAO Huili,SUN Haitong,BU Xianjiang
(School of Applied Sciences, Harbin University of Science and Technology, Harbin 150080, China)
Abstract:Since a class of specific differential equations called cellular neural networks with delays have been established, the discussion on existence and uniqueness of all kinds of solutions for this class of equation become important study contents, especially the study on almost periodic type solutions In order to study asymptotically almost functions,applications on a class of cellular neural networks with varying delays, the existence and uniqueness of asymptotically almost periodic solutions on this class of cellular neural networks are discussed by using Banach fixed point theorem and some knowledge on exponential dichotomy, combining with the research of almost periodic solutions on this class of cellular neural networks in some literature, and it is proved that this class of cellular neural networks has unique asymptotically almost periodic solution under some sufficient conditions
Keywords:cellular neural networks; asymptotically almost periodic solutions; exponential dichotomy; fixed point, delay
收稿日期: 2016-04-19
基金項(xiàng)目: 黑龍江省教育廳科學(xué)技術(shù)研究項(xiàng)目(12511110)
作者簡介:
姚慧麗(1970—),女,博士,教授,Email:huili_yao@sohu.com;
孫海彤(1992—) ,男,碩士研究生;
卜憲江(1991—),男,碩士研究生
0引言
1925年至1992年,概周期型函數(shù)有關(guān)理論相繼被提出[1-4],自提出以來已被數(shù)學(xué)研究
者廣泛應(yīng)用于各類微分方程中[5-7]。細(xì)胞神經(jīng)網(wǎng)絡(luò)是由歘和楊于1993年引進(jìn)的[8-9],自引進(jìn)以來,數(shù)學(xué)工作者已成功地將其應(yīng)用到各種領(lǐng)域,如信號處理、模式識別、靜態(tài)圖像處理、生物系統(tǒng)等[10-15],還有很多文獻(xiàn)對帶各類延遲的細(xì)胞神經(jīng)網(wǎng)絡(luò)的概周期解[16-18]和偽概周期解[19-20]進(jìn)行了研究。到目前,很少有文獻(xiàn)對這些方程的漸近概周期解進(jìn)行討論。本文的研究目的是對一類帶有可變延遲的細(xì)胞神經(jīng)網(wǎng)絡(luò)即方程(1)的漸近概周期解的存在唯一性問題進(jìn)行討論,由于漸近概周期函數(shù)是概周期函數(shù)加上擾動項(xiàng)構(gòu)成的,所以研究方程的漸近概周期解更具有一般性。
方程(1)如下:
x′i(t)=-ci(t)∫∞0hi(s)xi(t-s)ds+
∑nj=1aij(t)fj(xj(t-τij(t)))+
∑nj=1bij(t)∫∞0Kij(u)gj(xj(t-u))du+
Ii(t),i=1,2,…,n(1)
其中,n對應(yīng)著細(xì)胞元的數(shù)量,xi(t)表示在時(shí)間t時(shí)刻,第i個(gè)細(xì)胞元的狀態(tài)向量。ci(t)>0表示神經(jīng)網(wǎng)絡(luò)系統(tǒng)在沒有外部附加電壓,也沒有連通電源的情況下細(xì)胞元在t時(shí)刻恢復(fù)靜止?fàn)顟B(tài)的速率。aij(t)和bij(t)表示的是在t時(shí)刻連接的勢能。hi(s)≥0,Kij(u)和τij(t)分別表示泄漏延遲內(nèi)核,傳輸延遲內(nèi)核和傳輸延遲。Ii(t)表示一個(gè)在時(shí)刻t向細(xì)胞元外部輸入獨(dú)立的電流源。fj和gj表示的是信號傳輸?shù)募せ詈瘮?shù)。hi:R→[0,+∞)是一個(gè)連續(xù)函數(shù),ci:R→[0,+∞)是一個(gè)概周期函數(shù)。τij:R→[0,+∞),Ii,aij,bij: R→R都是漸近概周期函數(shù)。endprint
1預(yù)備知識
本文用R表示實(shí)數(shù)集,Rn表示所有n維實(shí)向量,C(R,Rn)表示的是從R到Rn的全體有界連續(xù)函數(shù)構(gòu)成的集合,并且(C(R,Rn),‖·‖∞)是一個(gè)Banach空間,其中‖f‖∞=supt∈R‖f(t)‖。若g是一個(gè)有界連續(xù)函數(shù),定義g+=supt∈R|g(t)|,g-=inft∈R|g(t)|。
設(shè)x=(x1,x2,…,xn)T∈Rn,|x|表示的是絕對值向量|x|=(|x1|,|x2|,…,|xn|)T,并且定義‖x‖=max1≤i≤n|xi|。且對兩個(gè)向量x=(x1,x2,…,xn)T,y=(y1,y2,…,yn)T,定義|x|≤|y|是指|xi|≤|yi|,i=1,2,…,n。
定義1函數(shù)f∈C(R,Rn)稱作是一個(gè)概周期函數(shù),如果它滿足:對任意的ε>0,存在一個(gè)lε>0,使得對任意a∈R,存在δ∈[α,α+lε]滿足‖f(·+δ)-f(·)‖∞<ε。用AP(R,Rn)來表示這類函數(shù)全體。
定義2函數(shù)f∈C(R,Rn)稱為是一個(gè)漸近概周期函數(shù),若它能夠表示成f=g+φ,其中g(shù)∈AP(R,Rn),φ∈C0(R,Rn)。這類函數(shù)全體用AAP(R,Rn)表示。其中
C0(R,Rn)={f∈C(R,Rn)|lim|t|→∞‖f(t)‖=0}
易知AP(R,Rn)和AAP(R,Rn)在上確界范數(shù)下都構(gòu)成Banach空間。
定義3令x∈Rn,Q(t)是定義在R上的一個(gè)n×n連續(xù)矩陣,稱線性系統(tǒng)
x′(t)=Q(t)x(t)(2)
滿足指數(shù)二分法,是指若存在正數(shù)k,α,射影P(P2=P),滿足
‖X(t)PX-1(s)‖≤ke-α(t-s)(t≥s)
‖X(t)(I-P)X-1(s)‖≤ke-α(s-t)(s≥t)
其中X(t)是方程(2)的基本矩陣解,并滿足X(0)=I。
定義4如果f:R→Rn是連續(xù)可微的 并且f和f′都是R上的漸近概周期函數(shù),則稱f是連續(xù)可微的漸近概周期函數(shù)。
引理1[21]假設(shè)Q(t)是一個(gè)概周期矩陣,g(t)∈AAP(R,Rn)。如果線性系統(tǒng)(2)滿足指數(shù)二分法,那么漸近概周期系統(tǒng)x′(t)=Q(t)x(t)+g(t)有唯一的漸近概周期解x(t),并且
x(t)=∫t-∞X(t)PX-1(s)g(s)ds-
∫+∞tX(t)(I-P)X-1(s)g(s)ds
引理2[21]若ci(t)是R上的概周期函數(shù),并且
M[ci(t)]=limT→∞1T∫t+Ttci(s)ds>0,i=1,2,…n
則線性系統(tǒng)x′(t)=diag(-c1(t),-c2(t),…,-cn(t))x(t)滿足指數(shù)二分法。
2主要結(jié)果
本部分是本文的主要研究結(jié)果。令
B={φ|φ=(φ1(t),…,φn(t))T}
其中φ是連續(xù)可微的漸近概周期函數(shù),其范數(shù)定義為
‖φ‖B=max{supt∈Rmax1≤i≤n|φi(t)|,supt∈Rmax1≤i≤n|φ′i(t)|}
則易知B是一個(gè)Banach空間。
定理1對方程(1)假設(shè)下列條件
(A1)對于每一個(gè)j∈{1,2,…,n},存在非負(fù)常數(shù) Lfj和 Lgj使得
|fj(u)-fj(v)|≤Lfj|u-v|,|gj(u)-gj(v)|≤Lgj|u-v|(u,v∈R)
(A2)對于每一個(gè)i,j∈{1,2,…,n}, Kij:[0,+∞)→R是連續(xù)的,且對確定的正數(shù)k,|Kij(t)|ekt在[0,+∞)上可積的;
(A3)對于每一個(gè)i∈{1,2,…,n},存在常數(shù)αi>0和 ξi>0,使得
-c-i∫∞0hi(v)dv+c+i∫∞0vhi(v)dv+ξ-1i∑nj=1a+ijLfjξj+ξ-1i∑nj=1b+ij∫∞t|Kij(u)|duLgjξj≤-αi
和
c-i∫∞0hi(v)dv-αi+c+i∫∞0hi(v)dv1-αic-i∫∞0hi(v)dv<1;
(A4)0<∫∞0hi(v)dv<∞,且對(A2)中確定的正數(shù)k有0<∫∞0hi(v)dv<∞,∫∞0vhi(v)ekvdv<∞,i=1,2,…,n
成立,那么方程(1)有唯一連續(xù)可微的漸近概周期解。
證明:設(shè)i(t)=ξ-1ixi(t),則可把方程(1)轉(zhuǎn)變成
′i(t)=-ci(t)∫∞0hi(s)i(t-s)ds+
ξ-1i∑nj=1aij(t)fj(ξjj(t-τij(t)))+
ξ-1i∑nj=1bij(t)∫∞0Kij(u)gj(ξjj(t-u))du+ξ-1iIi(t)=
-ci(t)∫∞0hi(s)dsi(t)+ci(t)∫∞0hi(s)∫tt-s′i(u)duds+
ξ-1i∑nj=1aij(t)fj(ξjj(t-τij(t)))
+ξ-1i∑nj=1bij(t)∫∞0Kij(u)gj(ξjj(t-u))du+ξ-1iIi(t),i=1,2,…,n
對于任意的φ∈B,考慮下列非線性微分方程
′i(t)=-ci(t)(∫∞0hi(s)ds)i(t)+
ci(t)∫∞0hi(s)∫tt-sφ′i(u)duds+
ξ-1i∑nj=1aij(t)fj(ξjφj(t-τij(t)))+
ξ-1i∑nj=1bij(t)∫∞0Kij(u)gj(ξjφj(t-u))duendprint
+ξ-1iIi(t),i=1,2,…,n,(3)
由于M[ci(t)∫∞0hi(s)ds]>0,i=1,2,…,n。根據(jù)引理2知線性系統(tǒng)
′i(t)=-ci(t)(∫∞0hi(s)ds)i(t) ,i=1,2,…,n
在R上滿足指數(shù)二分法,因此由引理1知方程(3)有一個(gè)漸近概周期解
xφ(t)=(xφ1(t),xφ2(t),…,xφn(t))T=
(∫t-∞e-∫tsc1(u)∫∞0h1(v)dvdu[c1(s)∫∞0h1(v)∫ss-vφ′1(u)dudv+
ξ-11∑nj=1a1j(s)fj(ξjφj(s-τ1j(s)))+
ξ-11∑nj=1b1j(s)∫∞0K1j(u)gj(ξjφj(s-u))du+ξ-11I1(s)]ds,…,
∫t-∞e-∫tscn(u)∫∞0hn(v)dvdu[cn(s)∫∞0hn(v)∫ss-vφ′n(u)dudv+
ξ-1n∑nj=1anj(s)fj(ξjφj(s-τnj(s)))+
ξ-1n∑nj=1bnj(s)∫∞0Knj(u)gj(ξjφj(s-u))du+ξ-1nIn(s)]ds)T(4)
定義一個(gè)映射T:T(φ)(t)=xφ(t), φ∈B。以下命題1至命題4證明T是從B到B的一個(gè)壓縮映射。
命題1假設(shè)(A1)和(A2)成立,且φ(·)∈AAP(R,R),則函數(shù)
∫∞0Kij(u)gj(φ(t-u))du∈AAP(R,R),
i,j=1,2,…,n
證明由于φ(t)∈AAP(R,R),所以對任意ε>0,存在一個(gè)相對稠子集Pε和一個(gè)有界子集Cε,使得|φ(s+τ)-φ(s)|<ε(τ∈Pε,s+τ,s∈R\Cε)。從而結(jié)合條件(A1)得
|gj(φ(s+τ))-gj(φ(s))|
≤Lgj|φ(s+τ)-φ(s)| 因此得gj(φ(t))∈AAP(R,R),從而gj(φ(t))=Xj1(t)+Xj2(t),其中 Xj1∈AP(R,R),Xj2∈C0(R,R) 由AP(R,R)的定義,對上述ε>0,存在一個(gè)正數(shù)l=l(ε),對長度為l的任一區(qū)間內(nèi)都存在一個(gè)數(shù)τ,使得 |Xj1(t+τ)-Xj1(t)|<ε1+∫∞0|Kij(u)|du,t∈R,i,j=1,2,…,n 因此有 |∫∞0Kij(u)Xj1(t+τ-u)du-∫∞0Kij(u)Xj1(t-u)du|≤ ∫∞0|Kij(u)||Xj1(t+τ-u)-Xj1(t-u)|du< ∫∞0|Kij(u)|duε1+∫∞0|Kij(u)|du<ε,t∈R, i,j=1,2,…,n 即得∫∞0Kij(u)Xj1(t-u)du∈AP(R,R),i=1,2,…,n。 下證lim|t|→∞|∫∞0Kij(u)Xj2(t-u)du|=0。 因?yàn)閄j2(t)連續(xù),且limt→∞|Xj2(t)|=0,故Xj2(t)在R上有界,即存在M1>0,使對任意t∈R,有|Xj2(t)|≤M1。由于limt→∞|Xj2(t)|=0,所以對任意ε>0,存在T1>0,當(dāng)|t|>T1時(shí),|Xj2(t)|<ε2。由條件(A2)知|Kij(t)|ekt在[0,+∞)是可積的,所以|Kij(t)|在[0,+∞)是可積的,故存在T2>0,使∫∞T2|Kij(u)|du<ε,并記 ∫T20|Kij(u)|du=G 令M=T1+T2,當(dāng)t<-M,u∈(0,+∞)時(shí),有t-u<-M<-T1 因此|Xj2(t-u)|<ε,于是|∫∞0Kij(u)Xj2(t-u)du|<ε∫∞0|Kij(u)|du。 當(dāng)t>M,u≤T2時(shí),t-u>M-T2=T1,因此|Xj2(t-u)|<ε。于是有 |∫∞0Kij(u)Xj2(t-u)du|=∫T20|Kij(u)Xj2(t-u)du+∫∞T2Kij(u)Xj2(t-u)du| ≤∫T20|Kij(u)||Xj2(t-u)|du+∫∞T2|Kij(u)||Xj2(t-u)|du ≤ε∫T20|Kij(u)|du+M1∫∞T2|Kij(u)|du <(G+M1)ε 因此有∫∞0Kij(u)Xj2(t-u)du∈C0(R,R)。從而得 ∫∞0Kij(u)gj(φ(t-u))du=∫∞0Kij(u)Xj1(t-u)du+∫∞0Kij(u)Xj2(t-u)du∈AAP(R,R) 命題2假設(shè)(A1)、(A2)和(A4)成立,且函數(shù)ci、τij、 Ii,aij,bij滿足引言中所述條件,φj∈B,則有 ci(t)∫∞0hi(s)∫tt-sφ′i(u)duds+ξ-1i∑nj=1aij(t)fj(ξjφj(t-τij(t))) +ξ-1i∑nj=1bij(t)∫∞0Kij(u)gj(ξjφj(t-u))du+ξ-1Ii(t)∈AAP(R,R),i=1,2,…,n 證明由于φj∈AAP(R,R),所以是一致連續(xù)的,從而對任意ε>0,存在>0,當(dāng)x1,x2∈R,|x1-x2|<時(shí),有|φj(x1)-φj(x2)|<ε。對上述的>0,由于τij∈AAP(R,R+),所以存在一個(gè)相對稠密的子集P1ε和一個(gè)有界子集C1ε,使得 |τij(t+s)-τij(t)|< (s∈P1ε,t,t+s∈R\C1ε) 而|t-τij(t+s)-(t-τij(t))|= |τij(t+s)-τij(t)|<,所以有
|φj(t-τij(t+s))-φj(t-τij(t))|<ε
由于φj∈AAP(R,R),所以對上述ε>0,存在一個(gè)相對稠密的子集P2ε和一個(gè)有界子集C2ε,使得|φj(t+s)-φj(t)|<ε(s∈P2ε,t,t+s∈R\C2ε)。
令Pε=P1ε∩P2ε, Cε=C1ε∪C2ε,則Pε為相對稠子集,Cε為有界子集,且有
|φj(t+s-τij(t+s))-φj(t-τij(t+s))|<ε
(s∈Pε,t-τij(t+s),t+s-τij(t+s)∈R\Cε)
因此有
|φj(t+s-τij(t+s))-φj(t-τij(t))|≤
|φj(t+s-τij(t+s))-φj(t-τij(t+s))+
φj(t-τij(t+s))-φj(t-τij(t))|≤
|φj(t+s-τij(t+s))-φj(t-τij(t+s))|+
|φj(t-τij(t+s))-φj(t-τij(t))|<
ε+ε=2ε
即得φj(t-τij(t))∈AAP(R,R),從而ξjφj(t-τij(t))∈AAP(R,R)。
下證fj(ξjφj(t-τij(t)))∈AAP(R,R)。
由于ξjφj(t-τij(t))∈AAP(R,R),所以對任意的ε>0,存在一個(gè)相對稠密的子集P3ε和一個(gè)有界子集C3ε,使得
|ξjφj(t+s-τij(t+s))-ξjφj(t-τij(t))|<ε
(s∈P3ε,t-τij(t),t+s-τij(t+s)∈R\C3ε)
結(jié)合條件(A1)有
|fj(ξjφj(t+s-τij(t+s)))-fj(ξjφj(t-τij(t)))|≤
Lfj|ξjφj(t+s-τij(t+s))-ξjφj(t-τij(t))|<
Lfj·ε
所以得
fj(ξjφj(t-τij(t)))∈AAP(R,R) ,i,j=1,2,…,n(5)
由條件(A4):0<∫∞0hi(v)dv<∞和0<∫∞0vhi(v)dv<∞,結(jié)合命題1,有
∫∞0hi(s)∫tt-sφ′i(u)duds=∫∞0hi(s)dsφi(t)-
∫∞0hi(s)dsφi(t-s)ds∈AAP(R,R)(6)
又條件(A1)和(A2)成立,所以命題1成立。結(jié)合式(5),(6)以及命題1,可得
ci(t)∫∞0hi(s)∫tt-sφ′i(u)duds+ξ-1i∑nj=1aij(t)fj(ξjφj(t-τij(t)))+
ξ-1i∑nj=1bij(t)∫∞0Kij(u)gj(ξjφj(t-u))du+ξ-1Ii(t)∈AAP(R,R),i=1,2,…,n
命題2證畢。
命題3在命題2的條件下,則式(4)中的xφ(t)和它的導(dǎo)數(shù)(xφ(t))′都是漸近概周期的。
證明令φi(s)=ci(s)∫∞0hi(v)∫ss-vφ′i(u)dudv+ξ-1i∑nj=1aij(s)fj(ξjφj(s-τij(s)))+ξ-1i∑nj=1bij(s)∫∞0Kij(u)gj(ξjφj(s-u))du+ξ-1iIi(s)
由命題2可知φi(s)∈AAP(R,R),因此可設(shè)φi(s)=φ1i(s)+φ2i(s),其中φ1i(s)∈AP(R,R),φ2i(s)∈C0(R,R)。于是對于1≤i≤n,有
∫t-∞e-∫tsci(u)∫∞0hi(v)dvduφi(s)ds=∫t-∞e-∫tsci(u)∫∞0hi(v)dvduφ1i(s)ds+∫t-∞e-∫tsci(u)∫∞0hi(v)dvduφ2i(s)ds
由文[19]知∫t-∞e-∫tsci(u)∫∞0hi(v)dvduφ1i(s)ds∈AP(R,R),下證
∫t-∞e-∫tsci(u)∫∞0hi(v)dvduφ2i(s)ds∈C0(R,R)
由于φ2i(s)有界,所以存在M′>0,使|φ2i(s)|
|∫t-∞e-∫tsci(u)∫∞0hi(v)dvduφ2i(s)ds-0|≤
∫t-∞e-∫tsci(u)∫∞0hi(v)dvdu|φ2i(s)|ds<ε∫t-∞e-∫tsGdu∫∞0hi(v)dvds<ε·∫t-∞e-G(t-s)∫∞0hi(v)dvds<εG∫∞0hi(v)dv
當(dāng)t>M時(shí),
|∫t-∞e-∫tsci(u)∫∞0hi(v)dvduφ2i(s)ds-0|≤∫t-∞|e-∫tsci(u)∫∞0hi(v)dvduφ2i(s)|ds
=∫-M1-∞|e-∫tsci(u)∫∞0hi(v)dvduφ2i(s)|ds+∫M1-M1|e-∫tsci(u)∫∞0hi(v)dvduφ2i(s)|ds
+∫tM1|e-∫tsci(u)∫∞0hi(v)dvduφ2i(s)|ds
≤εG∫∞0hi(v)dv+M′(∫M1-M1|e-∫tsci(u)∫∞0hi(v)dvdu|ds)+εG∫∞0hi(v)dv
≤εG∫∞0hi(v)dv+M′2M1εG∫∞0hi(v)dv+εG∫∞0hi(v)dv
所以xφi(t)∈AAP(R,R),i=1,2,…,n。
又因?yàn)?/p>
(xφi(t))′=[ci(t)∫∞0hi(v)∫tt-vφ′i(u)dudv+
ξ-1i∑nj=1aij(t)fj(ξjφj(t-τij(t)))
+ξ-1i∑nj=1bij(t)∫∞0Kij(u)gj(ξjφj(t-u))du+ξ-1iIi(t)]
-ci(t)∫∞0hi(v)dv∫t-∞e-∫tsci(u)∫∞0hi(v)dvdu[ci(s)∫∞0hi(v)∫ss-vφ′i(u)dudv
+ξ-1i∑nj=1aij(s)fj(ξjφj(s-τij(s)))
+ξ-1i∑nj=1bij(s)∫∞0Kij(u)gj(ξjφj(s-u))du+ξ-1iIi(s)]ds,i=1,2,…,n
再結(jié)合由命題1和命題2可得(xφi(t))′∈AAP(R,R),i=1,2,…,n。
因此由命題3可知T是從B到B的映射。
命題4上面定義的映射T是從B到B上的一個(gè)壓縮映射。
證明事實(shí)上,由式(4)及條件((A1)和(A3),對于φ,ψ∈B,有
|T(φ(t))-T(ψ(t))|=(|(T(φ(t))-T(ψ(t)))1|,…,
|(T(φ(t))-T(ψ(t)))n|)T
=(|∫t-∞e-∫tsc1(u)∫∞0h1(v)dvdu[c1(s)∫∞0h1(v)∫ss-v(φ′1(u)-ψ′1(u))dudv
+ξ-11∑nj=1a1j(s)fj(ξjφj(s-τ1j(s)))-fj(ξjψj(s-τ1j(s)))
+ξ-11∑nj=1b1j(s)∫∞0K1j(u)(gj(ξjφj(s-u))-gj(ξjψj(s-u)))du]ds|,…,
|∫t-∞e-∫tscn(u)∫∞0hn(v)dvdu[cn(s)∫∞0hn(v)∫ss-v(φ′n(u)-ψ′n(u))dudv
+ξ-1n∑nj=1anj(s)(fj(ξjφj(s-τnj(s)))-fj(ξjψj(s-τnj(s))))
+ξ-1n∑nj=1bnj(s)∫∞0Knj(u)(gj(ξjφj(s-u))-gj(ξjψj(s-u)))du]ds|)T
≤(∫t-∞e-∫tsc1(u)∫∞0h1(v)dvdu[c1(s)∫∞0h1(v)∫ss-v|φ′1(u)-ψ′1(u)|dudv
+ξ-11∑nj=1|a1j(s)|Lfjξj|φj(s-τ1j(s))-ψj(s-τ1j(s))|
+ξ-11∑nj=1|b1j(s)|∫∞0|K1j(u)|Lgjξj|φj(s-u)-ψj(s-u)|du]ds,…,
∫t-∞e-∫tscn(u)∫∞0hn(v)dvdu[cn(s)∫∞0hn(v)∫ss-v|φ′n(u)-ψ′n(u)|dudv
+ξ-1n∑nj=1|anj(s)|Lfjξj|φj(s-τnj(s))-ψj(s-τnj(s))|
+ξ-1n∑nj=1|bnj(s)|∫∞0|Knj(u)|Lgjξj|φj(s-u)-ψj(s-u)|du]ds)T
≤(∫t-∞e-∫tsc1(u)∫∞0h1(v)dvdu[c+1∫∞0vh1(v)dv+ξ-11∑nj=1(a+1jLfj
+b+1j∫∞0|K1j(u)|duLgj)ξj]ds‖φ(t)-ψ(t)‖B,…,
∫t-∞e-∫tscn(u)∫∞0hn(v)dvdu[c+n∫∞0vhn(v)dv+ξ-1n∑nj=1(a+njLfj
+b+nj∫∞0|Knj(u)|duLgj)ξj]ds‖φ(t)-ψ(t)‖B)T
≤(∫t-∞e-∫tsc1(u)∫∞0h1(v)dvdu(c1(s)∫∞0h1(v)dv-α1)ds,…,
∫t-∞e-∫tscn(u)∫∞0hn(v)dvdu(cn(s)∫∞0hn(v)dv-αn)ds)T‖φ(t)-ψ(t)‖B
≤(∫t-∞e∫tsc1(u)∫∞0h1(v)dvdu∫tsc1(s)∫∞0h1(v)dvds-
α1∫t-∞e-∫tsc1(u)∫∞0h1(v)dvduds,…,
∫t-∞e∫tscn(u)∫∞0hn(v)dvdu∫tscn(s)∫∞0hn(v)dvds-
αn∫t-∞e-∫tscn(u)∫∞0hn(v)dvduds)T‖φ(t)-ψ(t)‖B
≤(1-α1∫t-∞e-∫tsc+1∫∞0h1(v)dvduds,…,1-
αn∫t-∞e-∫tsc+n∫∞0hn(v)dvduds)T‖φ(t)-ψ(t)‖B
≤(1-α1c+1∫∞0h1(v)dv,…,1-αnc+n∫∞0hn(v)dv)T‖φ(t)-ψ(t)‖B(7)
且有
|T′(φ(t))-T′(ψ(t))|=(|(T′(φ(t))-T′(ψ(t)))1|,…,
|(T′(φ(t))-T′(ψ(t)))n|)T
=(|[c1(t)∫∞0h1(v)∫tt-v(φ′1(u)-ψ′1(u))dudv
+ξ-11∑nj=1a1j(t)(fj(ξjφj(t-τ1j(t)))-fj(ξjψj(t-τ1j(t))))
+ξ-11∑nj=1b1j(t)∫∞0K1j(u)(gj(ξjφj(t-u))-gj(ξjψj(t-u)))du]endprint
-c1(t)∫∞0h1(v)dv∫t-∞e-∫tsc1(u)∫∞0h1(v)dvdu[c1(s)∫∞0h1(v)∫ss-v(φ′1(u)-ψ′1(u))dudv
+ξ-11∑nj=1a1j(s)(fj(ξjφj(s-τ1j(s)))-fj(ξjψj(s-τ1j(s))))
+ξ-11∑nj=1b1j(s)∫∞0K1j(u)(gj(ξjφj(s-u))-gj(ξjψj(s-u)))du]ds|,…,
|[cn(t)∫∞0hn(v)∫tt-v(φ′n(u)-ψ′n(u))dudv
+ξ-1n∑nj=1anj(t)(fj(ξjφj(t-τnj(t)))-fj(ξjψj(t-τnj(t))))
+ξ-1n∑nj=1bnj(t)∫∞0Knj(u)(gj(ξjφj(t-u))-gj(ξjψj(t-u)))du]
-cn(t)∫∞0hn(v)dv∫t-∞e-∫tscn(u)∫∞0hn(v)dvdu[cn(s)∫∞0hn(v)∫ss-v(φ′n(u)-ψ′n(u))dudv
+ξ-1n∑nj=1anj(s)(fj(ξjφj(s-τnj(s)))-fj(ξjψj(s-τnj(s))))
+ξ-1n∑nj=1bnj(s)∫∞0Knj(u)(gj(ξjφj(s-u))-gj(ξjψj(s-u)))du]ds|)T
≤([c+1∫∞0vh1(v)dv+ξ-11∑nj=1(a+1jLfj+
b+1j∫∞0|K1j(u)|duLgj)ξj]‖φ(t)-ψ(t)‖B
+c+1∫∞0h1(v)dv∫t-∞e-∫tsc1(u)∫∞0h1(v)dvdu[c+1∫∞0vh1(v)dv
+ξ-11∑nj=1(a+1jLfj
+b+1j∫∞0|K1j(u)|duLgj)ξj]ds‖φ(t)-ψ(t)‖B,…,
[c+n∫∞0vhn(v)dv+ξ-1n∑nj=1(a+njLfj
+b+nj∫∞0|K1j(u)|duLgj)ξj]‖φ(t)-ψ(t)‖B
+c+n∫∞0hn(v)dv∫t-∞e-∫tscn(u)∫∞0hn(v)dvdu[c+n∫∞0vhn(v)dv+ξ-1n∑nj=1(a+njLfj
+b+nj∫∞0|Knj(u)|duLgj)ξj]ds‖φ(t)-ψ(t)‖B)T
≤(c-1∫∞0h1(v)dv-α1+c+1∫∞0h1(v)dv1-α1c+1∫∞0h1(v)dv,…,
c-n∫∞0hn(v)dv-αn
+c+n∫∞0hn(v)dv1-αnc+n∫∞0hn(v)dv)T‖φ(t)-ψ(t)‖B(8)
由條件(A3)可得0<1-αic+i∫∞0hi(v)dv<1以及
K=maxmax1≤i≤n1-αic+i∫∞0hi(v)dv,
max1≤i≤nc-i∫∞0hi(v)dv-αi+c+i∫∞0hi(v)dv1-αic+i∫∞0hi(v)dv<1
結(jié)合式(7)和式(8),可得‖T(φ(t)-T(ψ(t)))‖B≤K‖φ(t)-ψ(t)‖B,即映射T:B→B是一個(gè)壓縮映射。
因此由Banach不動點(diǎn)定理知映射T有一個(gè)唯一的不動點(diǎn)
x**=(x**1(t),x**2(t),…,x**n(t))T∈B,Tx**=x**
由式(3)和式(4)知x**滿足式(3)。因此方程(1)有一個(gè)唯一的連續(xù)可微的漸近概周期解x*=(ξ1x**1(t),ξ2x**2(t),…,ξnx**n(t))T。
參 考 文 獻(xiàn):
[1]BOHR H Zur Theorie Der Fastperiodischen Funktionne I[J]. Acta Math, 1925, 45: 19-127
[2]FRECHET M Les Functions Asymptotiquement Presqueperiodiques[J].Revue Sci, 1941, 79: 341-354
[3]EBERLEIN W F Abstract Ergodic Theorems and Weak Almost Periodic Functions[J]. Trans Amer Math Soc, 1949, 67: 217-240
[4]ZHANG C Pseudo Almost Periodic Functions and Their Applications[D]. Canada:University of Western Ontario 1992, 24: 23-29
[5]DADS E A, ARINO O Exponential Dichotomy and Existence of Pseudo Almost Periodic Solutions of Some Differential Equations[J]. Nonlinear Analysis, 1996,27(4): 369-386
[6]DADS E A, EZZINBI K, ARINO O Pseudo Almost Periodic Solutions forSome Differential Equations in A Banach Space[J]. Nonlinear Analysis, 1997,28: 1141-1155
[7]姚慧麗,王健偉。一類隨機(jī)積分-微分方程的均方漸近概周期解[J].哈爾濱理工大學(xué)學(xué)報(bào),2014,19(6):118-122endprint
[8]CHUA L O, YANG L Cellular Neural Network[J]. Circuits Syst, 1988, 35 (10): 1273-1290
[9]CHUA L O, YANG L Cellular Neural Network[J]. Circuits Syst, 1988, 35(10):1257-1272
[10]WEN GANG, XIAO HONG Attractors of Discrete Cellular Neural Networks[J]. Artificial Intelligence and Computational Intelligence,2011, 7004: 293-299
[11]LIU B, HUANG L Existence and Exponential Stability of Almost Periodic Solutions for Cellular Neural Networks with Timevarying Delays[J]. Phys Lett A, 2005,341(1):135-144
[12]LIu B, HUANG L Positive Almost Periodic Solutions for Recurrent Neural Networks [J]. Nonlinear Anal Real Word Appl, 2008,9:830-841
[13]LU W, CHEN T Global Exponential Stability of Almost Periodic Solutions for a Large Class Delayed Dynamical Systems[J]. Sci China Ser A Math, 2005,48(8):1015-1026
[14]HUANG Z, MOHAMAD S, FENG C New Results on Exponential Attractivity of Multiple Almost Periodic Solutions of Cellular Neural Networks with Timevarying Delays[J]. Math Comput Model ,2010,52(10):1521-1531
[15]CHUA LO, ROSKA T Cellular Neural Networks with Nonlinear and Delaytype Template Elements [J]//In:Proceeding 1990 IEEE International Workshop on Cellular Neural Networks and Their Applications,:12-25
[16]CHEN A, CAO J Almost Periodic Solution of Shunting Inhibitory CNNs with Delays[J]. Phys Lett, 2002, 41: 221-236
[17]ZHANG HONG Existence and Stability of Almost Periodic Solutions for CNNs with Continuously Distributed Leakage Delays[J]. Neural Comput and Applic, 2014, 24: 1135-1146
[18]XU YANLI New Results on Almost Periodic Solutions for CNNs with Timevarying Leakage Delays[J]. Neural Comput and Applic, 2014, 25: 1293-1302
[19]LIU BINGWEN, CEMIL T Pseudo Almost Periodic Solutions for CNNs with Leakage Delays and Complex Deviating Argument[J]. Neural Comput & Applic, 2015, 26(2): 429-435
[20]LIU BINGWEN, CEMIL T Pseudo Almost Periodic Solutions for CNNs with Leakage Delays and Complex Deviating Argument[J]. Neural Comput & Applic, 2015, 26(2): 429-435
[21]張娜。兩類擾動系統(tǒng)的漸近概周期解[D]. 哈爾濱:哈爾濱理工大學(xué),2015:1-35
(編輯:王萍)endprint