吳春雪
高階脈沖變時滯BAM神經(jīng)網(wǎng)絡(luò)的周期解
吳春雪
(煙臺大學(xué)數(shù)學(xué)與信息科學(xué)學(xué)院,山東煙臺264005)
神經(jīng)網(wǎng)絡(luò)的諸多功能主要體現(xiàn)在其動力學(xué)特征中,而周期解問題則是其動力學(xué)行為研究中很重要的一部分.許多情況下,考慮神經(jīng)網(wǎng)絡(luò)的脈沖效應(yīng)是必要而具有實際價值的.本文利用重合度理論中的Gaines-Mawhin延拓定理和微分不等式技巧,研究一類具脈沖干擾的高階BAM神經(jīng)網(wǎng)絡(luò)模型的周期解問題,在要求激活函數(shù)有界的前提下,得到其周期解存在的充分條件.
重合度理論;BAM神經(jīng)網(wǎng)絡(luò);周期解;脈沖
1987年和1988年由Kosko提出雙向聯(lián)想記憶(BAM)神經(jīng)網(wǎng)絡(luò)[1-2],這類神經(jīng)網(wǎng)絡(luò)在圖像信號處理、自動控制、人工智能、聯(lián)想記憶、解最優(yōu)化等方面具有廣泛的應(yīng)用.關(guān)于BAM神經(jīng)網(wǎng)絡(luò)模型解的存在性及穩(wěn)定性問題,目前已經(jīng)有不少研究成果,見文獻[3-11].
由于高階神經(jīng)網(wǎng)絡(luò)有著許多優(yōu)良的性質(zhì),強收斂率和高儲存量等,吸引了許多學(xué)者的關(guān)注.大多研究BAM神經(jīng)網(wǎng)絡(luò)的文獻中,或只考慮時滯模型[12],或只考慮脈沖模型[13],本文在已有文獻的基礎(chǔ)上,利用重合度理論和微分不等式的技巧對高階脈沖時滯BAM神經(jīng)網(wǎng)絡(luò)進行研究,得到了周期解的存在性定理.
引理1(Gaines-Mawhin延拓定理)[14]設(shè)X是Banach空間,L為指標為零的Fredholm算子,Ω為X中有界開集,N:Ω→X連續(xù)映射且在ˉΩ是L-緊的,如果下列條件成立:
(1)Lx≠λNx,?x∈?Ω∩DomL,?λ∈(0,1);
(2)QNx≠0,?x∈?Ω∩KerL;
(3)deg{JQN,Ω∩KerL,0}≠0,
則方程Lx=Nx在DomL∩ˉΩ中至少存在一個解.對于任意非負整數(shù)q,令
令X={u∈C[0,ω;t1,t2,…,tq]|u(t)=u(t+ω)},Z=X×?(n+m)×(q+1),
下面介紹符號標記.
我們考慮如下網(wǎng)絡(luò)模型
其中模型的神經(jīng)網(wǎng)絡(luò)方面的意義見文獻[15]和[16].
假設(shè)下面條件成立:
(H1)函數(shù)fjil(u),gijl(u)滿足Lipschitz條件,即存在常數(shù)>0,使得
對于任意的u1,u2∈?,u1≠u2,i=1,2,…,n,j=1,2,…,m;
(H2)存在一個正整數(shù)q,使得
(H3)ai(t)>0,bj(t)>0,ci(t),dj(t),pjil(t),qijl(t)和τjil(t),σijl(t)都是T周期函數(shù),且時滯0≤τjil(t)≤τ,0≤σijl(t)≤σ(i=1,2,…,n,j=1,2,…,m,l=1,2,…,s);
(H4)存在正常數(shù)Mijl,Nijl使得
|fijl(u)|≤Mijl,|gijl(u)|≤Nijl,?u∈?,i=1,2,…,n,j=1,2,…,m,l=1,2,…,s.
定理1假設(shè)條件(H1)~(H4)成立,則系統(tǒng)(1)至少有一個T-周期解.
證明取L:DomL∩X→Z,Lu=(u',Δu(t1),…,Δu(tq),0),
考慮相應(yīng)算子方程Lx=λNx,λ∈(0,1),有
假設(shè)u(t)=(x1(t),x2(t),…,xn(t),y1(t),y2(t),…,ym(t))T∈X是系統(tǒng)(2)的對于某個λ∈(0,1)的解,在區(qū)間[0,ω]上積分式(2),于是獲得
整理得
令ξi∈[0,ω](≠tk),k=1,…,q,使得,i=1,2,…,n.則利用式(3)和H¨oder不等式,有
根據(jù)式(2)和(3)有
方程(2)兩邊同乘xi(t)且在[0,ω]上積分,由于
同理可得‖yj‖2≤.
將式(6)代入式(4)得
取Ω={u(t)∈X|‖u‖<H*},顯然滿足引理1的條件(1),當u∈?Ω∩?n+m,u是?n+m中常向量且,則QNu=(E1,E2,…,En,En+1,…,En+m)T.
因此,對于u(t)∈?Ω∩KerL,滿足引理1中的條件(2).
定義一個連續(xù)函數(shù)H:DomL×[0,1]→X滿足H(μ,u)=-μx+(1-μ)QNu,這里u(t)是?n+m中的常向量且μ∈[0,1],因此‖H(x1,…,xn,y1,…,ym,μ)‖>0.
選擇J為恒等映射,有deg{JQN,Ω∩KerL,0}=deg{-u,Ω∩KerL,0}≠0滿足引理1中的條件(3).
于是Ω滿足引理1中所有條件,方程(1)至少有一個ω周期解.
參考文獻:
[1]Kosko B.Adaptive bi-directional associative memories[J].Appl Opt,1987,26(23):4947-4960.
[2]Kosko B.Bi-directional associative memories[J].IEEE Trans Syst Man Cybernet,1988,18:49-60.
[3]Wang Qi,Wang Rui,Zheng Zhaoyue.2Nweighted pseudo almost periodic solutions for CNNs with variable and distributed delays[J].Mathematica Applicata,2014,27(1):73-81.
[4]Agarwal R P,Andrade B D,Cuevas C.Weighted pseudo-almost periodic solutions of a class of semilinear fractional differential equations[J].Nonlinear Ana,2010,11:3532-3554.
[5]Chen Xiaoxing,Hu Xiangyang.Weighted pseudo almost periodic solutions of neutral functional differential equations[J].Nonlinear Analy,2011,12:601-610.
[6]Ding Wei,Wang Linghai.2Nalmost periodic attractors for Cohen-Grossberg-type BAM neural networks with variable coefficients and distributed delays[J].J Math Anal Appl,2011,373:322-342.
[7]Gui Zhanji,Ge Weigao.Periodic solutions of nonautonomous cellular neural networks with impulses[J].Chaos,Solitons and Fractals,2007,32(5):1760-1771.
[8]Li Yongkun.Global exponential stability of BAM neural networks with delays and impulses[J].Chaos,Solitons and Fractals,2005,24:279-285.
[9]Song Qiankun,Wang Zidong.An analysis on existence and global exponential stability of periodic solutions for BAM neural networks with time-varying delays[J].Nonlinear Analysis,2007:1224-1234.
[10]Lou Xuyang,Cui Baotong,Wu Wei.On global exponential stability and existence of periodic solutions for BAM neural networks with distributed delays and reaction-diffusion terms[J].Chaos,Solitons and Fractals,2008,36:1044-1054.
[11]Li Yongkun.Existence and stability of periodic solution for BAM neural networks with distributed delays[J].Applied Mathematics and Computation,2004:847-862.
[12]Gopalsamy K,He Xuezhong.Delay-independent stability in bi-directional associative memory networks[J].IEEE Transactions on Neural Networks,1994,5:998-1002.
[13]Gui Zhanji,Yang Xiaosong,Ge Weigao.Periodic solution for nonautonomous bidirectional associative memory neural networks with impulses[J].Neurocomputing,2007,70:2517-2527.
[14]Gaines R E,Mawhin J L.Coincidence Degree and Nonlinear Differential Equations[M].New York:Springer Verlag,1977: 1-50.
[15]Ho W C,Liang Jinling,Lam J.Global exponential stability of impulsive high-order BAM neural networks with time-varying delays[J].Neural Networks,2006,19:1581-1590.
[16]Zhang Jie,Gui Zhanji.Periodic solutions of nonautonomous cellular neural networks with impulses and delays[J].Nonlinear Analysis:RWA,2009,10:1861-1903.
Periodic Solution of Impulsive High-order BAM Neural Networks with Time-varying Delays
WU Chun-xue
(School of Mathematics and Information Science,Yantai University,Yantai 264005,China)
The information processing function of neural networks mostly reflects in its dynamic characteristics.The periodic solution problem is one of the most important parts in the research of neural network dynamic actions in many cases.It is necessary and practically valuable to consider the impulse effect of neural networks.In this paper,by using the continuation theorem of Mawhin’s coincidence degree theory and differential inequalities,sufficient conditions are obtained for the existence of periodic solution of higher-order BAM neural networks with variable delays and impulses under the requirement of boundedness of involved functions.
coincidence degree;BAM neural network;periodic solution;impulse
O175
A
(責(zé)任編輯 李春梅)
1004-8820(2015)03-0157-05
10.13951/j.cnki.37-1213/n.2015.03.001
2014-09-25
吳春雪(1976-),女,黑龍江雙城人,講師,博士,研究方向:微分方程.