易書明 蹇繼貴
(三峽大學(xué) 理學(xué)院, 湖北 宜昌 443002)
?
基于憶阻的脈沖BAM神經(jīng)網(wǎng)絡(luò)的拉格朗日穩(wěn)定性
易書明蹇繼貴
(三峽大學(xué) 理學(xué)院, 湖北 宜昌443002)
摘要:本文研究了一類帶有時(shí)滯的憶阻脈沖BAM(Bidirectional Associative Memory)神經(jīng)網(wǎng)絡(luò)的Lagrange穩(wěn)定性.利用Lyapunov函數(shù)和不等式方法,得到時(shí)滯憶阻脈沖BAM神經(jīng)網(wǎng)絡(luò)的Lagrange穩(wěn)定性的充分條件,根據(jù)系統(tǒng)自身參數(shù)給出了其全局指數(shù)吸引集的估計(jì).最后,通過數(shù)值實(shí)例驗(yàn)證了理論的正確性.
關(guān)鍵詞:BAM神經(jīng)網(wǎng)絡(luò);憶阻器;脈沖;拉格朗日穩(wěn)定性;李雅普諾夫函數(shù);不等式
雙向聯(lián)想記憶(Bidirectional Associative Memory, BAM)神經(jīng)網(wǎng)絡(luò)模型最早由Kosko[1-3]提出,這類網(wǎng)絡(luò)在模式識(shí)別、信號處理和人工智能等方面得到廣泛應(yīng)用.目前對BAM神經(jīng)網(wǎng)絡(luò)的動(dòng)力學(xué)行為如平衡點(diǎn)的存在性、唯一性和全局穩(wěn)定性的研究出現(xiàn)了大量成果[4-11].
眾所周知,脈沖現(xiàn)象影響著神經(jīng)網(wǎng)絡(luò)的穩(wěn)定性[12-13],脈沖的存在意味著狀態(tài)軌跡不會(huì)一成不變.在文獻(xiàn)[14]中,關(guān)治洪教授等討論了時(shí)滯脈沖Hopfield神經(jīng)網(wǎng)絡(luò)的平衡點(diǎn)的存在性、唯一性和全局穩(wěn)定性.同時(shí),關(guān)于時(shí)滯脈沖神經(jīng)網(wǎng)絡(luò)的漸近或指數(shù)穩(wěn)定也被廣泛研究[15-16].
20世紀(jì)70年代,蔡少棠教授[17]從邏輯和公理的觀點(diǎn)指出,自然界應(yīng)該還存在一個(gè)電路元件,它表示磁通與電荷的關(guān)系,這就是憶阻器.隨著科學(xué)的發(fā)展,惠普公司在2008年做出了納米憶阻器,引起全球?qū)涀柩芯康膹V泛關(guān)注[18-19].憶阻器是模擬人工神經(jīng)網(wǎng)絡(luò)突觸的最佳原件,因此,許多研究者對基于憶阻的神經(jīng)網(wǎng)絡(luò)進(jìn)行了研究[20-23].在文獻(xiàn)[24-25]中,吳愛龍和曾志剛教授考慮了一類含有憶阻突觸和多重滯后的神經(jīng)網(wǎng)絡(luò),研究了它的有界性.在文獻(xiàn)[26]中,張國東博士等研究了一類憶阻遞歸神經(jīng)網(wǎng)絡(luò)的Lagrange穩(wěn)定性.而對于帶有時(shí)滯的憶阻脈沖BAM神經(jīng)網(wǎng)絡(luò)的Lagrange穩(wěn)定性的研究成果還沒有發(fā)現(xiàn),因此,本文建立一種新的時(shí)滯憶阻脈沖BAM神經(jīng)網(wǎng)絡(luò),并運(yùn)用不等式技巧討論其Lagrange穩(wěn)定性和全局指數(shù)吸引集.
考慮如下憶阻脈沖BAM神經(jīng)網(wǎng)絡(luò):
(1)
假設(shè)系統(tǒng)(1)的初始條件為
(2)
其中φi(s),ψj(s)是定義在[-τ,0]上的連續(xù)函數(shù).令
考慮如下兩種函數(shù)集合B={p(x)|p(x)∈C(R,R),?ξ>0,|p(x)|≤ξ,?x∈R},S={p(x)|p(x),p(y)∈C(R,R),?ζ>0,|p(x)-p(y)|≤ζ|x-y|,?x,y∈R},令
定義1[11]稱系統(tǒng)(1)是一致有界的.若?H>0,?K=K(φ,ψ)>0,使得‖(xT(t),yT(t))‖≤K(φ,ψ)對所有(φ,ψ)∈CH,t≥0成立.
定義2[27]稱系統(tǒng)(1)是Lagrange全局指數(shù)穩(wěn)定的,若存在正定徑向無界的函數(shù)V(x,y),函數(shù)K(φ,ψ)∈C,l>0,α>0,使得對系統(tǒng)(1)的任意解x(t)=x(t;φ,ψ),y(t)=y(t;φ,ψ),V(x,y)>l,t≥0,有
(3)
緊集Ω:={x∈Rn,y∈Rm|V(x,y)≤l}稱為系統(tǒng)(1)的全局指數(shù)吸引集.
引理1[27]設(shè)G∈C([t,+∞],R),存在正常數(shù)α和β使得
(4)
那么有
(5)
1主要成果
證:構(gòu)造正定徑向無界的Lyapunov函數(shù)
當(dāng)t≠tk時(shí),
由引理1可得
當(dāng)t=tk時(shí)
則有
綜上所述,對任意t>0有
由定義2知,系統(tǒng)(1)是Lagrange全局指數(shù)穩(wěn)定的,且Ω1是(1)的全局指數(shù)吸引集.
證:構(gòu)造正定徑向無界的Lyapunov函數(shù)
當(dāng)t≠tk時(shí)
由上式可以得到
由引理2可以得出
其中λ是方程λ=L2-L3eλτ的唯一正根.
當(dāng)t=tk時(shí),
綜上所述,對任意t>0有
由定義2知,系統(tǒng)(1)是Lagrange全局指數(shù)穩(wěn)定的,且Ω2是(1)的全局指數(shù)吸引集.
注1:在本文的條件下,定理1和定理2通過選取的特定Lyapunov函數(shù)得到的結(jié)果與時(shí)滯無關(guān).因此,無論有限時(shí)滯,還是無限時(shí)滯,都不會(huì)影響定理的正確性.
2仿真實(shí)例
同時(shí),取初始條件x1(0)=0.7,x2(0)=1,y1(0)=1.2,y2(0)=0.9,圖1表示x1(t),x2(t),y1(t),y2(t)隨時(shí)間t變化的狀態(tài)圖,圖2~5顯示系統(tǒng)(1)分別在三維相空間內(nèi)的界估計(jì).
圖1 x1(t),x2(t),y1(t),y2(t)隨時(shí)間t變化的狀態(tài)圖
圖2 系統(tǒng)(1)在坐標(biāo)系(x1,x2,y1)內(nèi)的界估計(jì)
圖3 系統(tǒng)(1)在坐標(biāo)系(x1,x2,y2)內(nèi)的界估計(jì)
圖4 系統(tǒng)(1)在坐標(biāo)系(x1,y1,y2)內(nèi)的界估計(jì)
圖5 系統(tǒng)(1)在坐標(biāo)系(x2,y1,y2)內(nèi)的界估計(jì)
3結(jié)語
本文運(yùn)用Lyapunov函數(shù)和不等式方法研究了時(shí)滯脈沖的憶阻BAM神經(jīng)網(wǎng)絡(luò)的Lagrange穩(wěn)定性,得到了Lagrange全局指數(shù)穩(wěn)定的充分條件,并對其全局指數(shù)吸引集進(jìn)行界估計(jì).最后,通過數(shù)值實(shí)驗(yàn)驗(yàn)證了理論的正確性.
參考文獻(xiàn):
[1]Kosko B.Adaptive Bidirectional Associative Memories[J].Appl.Opt.,1987,26:4947-4960.
[2]Kosko B.Bidirectional Associative Memories[J].IEEE Trans.Syst.Man Cybern,1988,18:49-60.
[3]Kosko B.Neural Networks and Fuzzy Systems: A Dynamical Systems Approach to Machine Intelligence[J].prentice hall,1992.
[4]Wang B X,Jian J G.Stability and Hopf Bifurcation Analysis on a Four-neuron Bam Neural Network with Distributed Delays[J].Commun Non-linear Sci.Numer.Simulat.,2010,15:189-204.
[5]Jian J G, Wang B X.Stability Analysis in Lagrange Sense for a Class of Bam Neural Networks of Neutral Type with Multiple Time-varying Delays[J].Neurocomputing,2015,149:930-939.
[6]Jian J G,Wang B X.Global Lagrange Stability for Neutral-type Cohen-grossberg BAM Neural Networks with Mixed Time-varying Delays[J].Math.Comput.Simul.,2015,116:1-25.
[7]Zhao Z H,Jian J G.Attracting and Quasi-invariant Sets for Bam Neural Networks of Neutral-type with Time-varying and Infinite Distributed Delays[J].Neurocomputing,2014, 140:265-272.
[8]Zhao Z H, Jian J G.Positive Invariant Sets and Global Exponential Attractive Sets of Bam Neural Networks with Time-varying and Infinite Distributed Delays[J].Neurocomputing,2014, 142:447-457.
[9]Zhao Z H,Jian J G,Wang B X.Global Attracting Sets for Neutral Type Bam Neural Networks with Time-varying and Inifnite Distributed Delays[J].Nonlinear Analysis: HS.,2015,15:63-73.
[10] Jian J G,Wang B X.Global Lagrange Stability for Neutral-type Cohen-Grossberg BAM Neural Networks with Mixed Time-varying Delays[J].Math.a(chǎn)nd Comput.Simulat.,2015,116:1-25.
[11] Li L L,Jian J G.Exponential P-convergence Analysis for Stochastic BAM Neural Networks with Time-varying and Infinite Distributed Delays[J].Appl.Math.Comput.,2015,266:860-873.
[12] Li X D.Existence and Global Exponential Stability of Periodic Solution for Impulsive Cohen-grossberg-type BAM Neural Networks with Continuously Distributed Delays[J].Appl.Math.Comput,2009,215(1):292-307.
[13] Zhang Y,Sun J.Stability of Impulsive Neural Networks with Time Delays[J].Phys.Lett.A,2005,348(1):44-50.
[14] Guan Z H,Chen G R.On Delayed Impulsive Hopfield Neural Networks[J].Neural Networks,1999,12:273-280.
[15] Song Q K,Cao J D.Impulsive E?ects on Stability of Fuzzy Cohen-grossberg Neural Networks with Time-varying Delays[J].IEEE Trans.Syst.Man Cybern.,Part B: Cybern.,2007,37(3):733-741.
[16] Zhu Q X, Cao J D.Stability Analysis of Markovian Jump Stochastic Bam Neural Networks with Impulse Control and Mixed Time Delays[J].IEEE Trans.Neural Netw.Learning Syst.,2012,23(3):467-479.
[17] Chua L O.Memristor-the Missing Circuit Element[J].IEEE Trans.Circ.Theory,1971,18(5):507-519.
[18] Corinto F,Ascoli A,Gilli M.Nonlinear Dynamics of Memristor Oscillators[J].IEEE Trans.Circ.Syst.I,Reg.Papers,2011,58(6):1323-1336.
[19] Ho Y,Huang G,Li P.Dynamical Properties and Design Analysis for Nonvolatile Memristor Memories[J].IEEE Trans.Circ.Syst.I, Reg.Papers,2011,58(4):724-736.
[20] Chen J J,Zeng Z G,Jiang P.Global Mittag-Leffier Stability and Synchronization of Memristor-based Fractional-order Neural Networks[J]. Neural Netw.,2014,51:1-8.
[21] Cai Z,Huang L.Functional Differential Inclusions and Dynamic Behaviors for Memristor-based BAM Neural Networks with Time-varying Delays[J]. Commun.Nonlinear Sci.Numer.Simul.,2014,19: 1279-1300.
[22] Chen J J,Zeng Z G,Jiang P.On the Periodic Dynamics of Memristor-based Neural Networks with Time-varying Delays[J]. Info.Sci.,2014,279: 358-373.
[23] Jiang M H,Mei J,Hu J H.New Results on Exponential Synchronization of Memristor-based Chaotic Neural Networks[J]. Neurocomputing,2015,156:60-67.
[24] Wu A L,Zeng Z G.Lagrange Stability of Neural Networks with Memristive Synapses and Multiple Delays[J].Info.Sci.,2012,280:135-151.
[25] Wu A I,Zeng Z G.Lagrange Stability of Memristive Neural Networks With Discrete and Distributed Delays[J]. Ieee Transactions on Neural Networks and Learning Systems, 2014,25(4):690-703.
[26] Zhang G D,Shen Y,Xu C J.Global Exponential Stability in a Lagrange Sense for Memristive Recurrent Neural Networks with Time-varying Delays[J].Neurocomputing 2015,149:1330-1336.
[27] Liao X X,Luo Q,Zeng Z G,et al. Global Exponential Stability in Lagrange Sense for Recurrent Neural Networks with Time Delays[J],Nonlinear Anal.:RWA.,2008,9:1535-1557.
[28] Jian J G,Kong D M,Luo H G.Exponential Stability of Differential Systems with Separated Variables and Time Delays[J].Journal of Central South University,2005,36(2):282-287.
[責(zé)任編輯張莉]
DOI:10.13393/j.cnki.issn.1672-948X.2016.03.022
收稿日期:2016-03-08
基金項(xiàng)目:國家自然科學(xué)基金(61273183,61304162,61174216)
通信作者:蹇繼貴(1965-),男,教授,博士,主要從事系統(tǒng)的穩(wěn)定性,神經(jīng)網(wǎng)絡(luò)理論,非線性系統(tǒng)控制等研究.E-mail:jiguijian@ctgu.edu.cn
中圖分類號:O231.2
文獻(xiàn)標(biāo)識(shí)碼:A
文章編號:1672-948X(2016)03-0098-06
Lagrange Stability for Memristive BAM Neural Networks with Impulse
Yi Shuming Jian Jigui
(College of Science, China Three Gorges Univ., Yichang 443002, China)
AbstractThis paper investigates Lagrange stability for a class of memristive BAM impulse neural networks with multiple time-varying delays and finds the global exponential attractive sets of it.By applying inequality techniques and Lyapunov function, some easily verifiable delay-independent criteria for the Lagrange stability and global exponential attractive sets of memristive BAM impulse neural networks are obtained by constructing appropriate Lyapunov functions. Finally, an example with numerical simulations is given to illustrate the results obtained.
KeywordsBAM neural networks;memristor;impulse;Lagrange stability;Lyapunov function;inequality