崔世維,蓋明久,劉孝磊
(海軍航空工程學(xué)院a.研究生管理大隊;b.基礎(chǔ)部,山東煙臺264001)
分?jǐn)?shù)階Hopfield型神經(jīng)網(wǎng)絡(luò)的全局漸近穩(wěn)定性
崔世維a,蓋明久b,劉孝磊b
(海軍航空工程學(xué)院a.研究生管理大隊;b.基礎(chǔ)部,山東煙臺264001)
研究了分?jǐn)?shù)階Hopfield型神經(jīng)網(wǎng)絡(luò)的全局漸近穩(wěn)定性,通過LMI方法得到了一種實現(xiàn)系統(tǒng)全局漸近穩(wěn)定性的LMI形式條件,通過實例仿真驗證了結(jié)論的正確性。
分?jǐn)?shù)階;神經(jīng)網(wǎng)絡(luò);漸近穩(wěn)定性;LMI
隨著對分?jǐn)?shù)階微積分的認(rèn)識越來越深刻,分?jǐn)?shù)階微積分在多個學(xué)科有廣泛應(yīng)用,如物理、化學(xué)、生物和圖像處理等,并且取得了較好的研究成果,凸顯了其獨特的優(yōu)勢和不可替代性,其理論和應(yīng)用研究引起了廣泛重視,已經(jīng)成為當(dāng)前國際上的一個研究熱點[1-7]。
眾所周知,小到一個具體的控制系統(tǒng),大至社會系統(tǒng)、金融系統(tǒng)、生態(tài)系統(tǒng),總是在各種偶然的或持續(xù)的干擾下進行的,保持系統(tǒng)的穩(wěn)定性就顯得至關(guān)重要。因此,研究分?jǐn)?shù)階系統(tǒng)也必不可免地面臨著對穩(wěn)定性的研究。目前,關(guān)于分?jǐn)?shù)階系統(tǒng)穩(wěn)定性的研究結(jié)果主要涉及分?jǐn)?shù)階系統(tǒng)的有限時穩(wěn)定性、魯棒穩(wěn)定性、Mittag-Leffler穩(wěn)定性等[8-11]。然而,由于分?jǐn)?shù)階微積分定義的不統(tǒng)一以及定義本身的復(fù)雜性,導(dǎo)致對分?jǐn)?shù)階系統(tǒng)穩(wěn)定性的研究面臨著眾多困難。
另一方面,由于神經(jīng)網(wǎng)絡(luò)良好的非線性映射能力、自學(xué)習(xí)適應(yīng)能力和并行信息處理能力,為解決不確定非線性系統(tǒng)的建模和控制提供了新思路。隨著分?jǐn)?shù)階微積分的發(fā)展,人們將分?jǐn)?shù)階微積分引入到神經(jīng)網(wǎng)絡(luò),建立了分?jǐn)?shù)階神經(jīng)網(wǎng)絡(luò),并取得了一定的研究成果。文獻[12]基于整數(shù)階Hopfield神經(jīng)網(wǎng)絡(luò)模型提出了分?jǐn)?shù)階Hopfield型神經(jīng)網(wǎng)絡(luò)模型,并引起了人們對分?jǐn)?shù)階Hopfield型神經(jīng)網(wǎng)絡(luò)的穩(wěn)定性及特殊結(jié)構(gòu)的動力行為等方面的廣泛關(guān)注[13-15]。
本文研究了分?jǐn)?shù)階Hopfield型神經(jīng)網(wǎng)絡(luò)的全局漸近穩(wěn)定性,得到了一種LMI形式的充分條件,并通過實例仿真驗證了結(jié)論的正確性。
定義1:[1-2]如果 f(t)∈C1(0,+∞),那么稱
為函數(shù) f()
定義2:[1-2]如果 f(t)∈C1(0,+∞),那么稱
為函數(shù) f(t)的α階Caputo型分?jǐn)?shù)階導(dǎo)數(shù)。
為簡便起見,下文將α階Caputo導(dǎo)數(shù)記為Dα。由文獻[1]知,Dα的Laplace變換式為
定義3:[1-2]Mittag-Leffler函數(shù)的單參數(shù)和雙參數(shù)分別定義為:
由文獻[1]知,雙參數(shù)Mittag-Leffler函數(shù)的Laplace變換式為:
考慮如下分?jǐn)?shù)階神經(jīng)網(wǎng)絡(luò):
式(1)的向量形式為:
為研究分?jǐn)?shù)階系統(tǒng)的穩(wěn)定性,需要下面假設(shè)及引理。
假設(shè)1:激勵函數(shù) fi(·)(i=1,2,…,n)有界,且滿足Lipschitz條件,即存在Lipschitz常數(shù)li使得
引理1:[1-2]Caputo導(dǎo)數(shù)Dα具有以下性質(zhì):
1)Dαc=0,這里c為常數(shù);
2)Dα(μf(t)+νg(t))=μDαf(t)+νDαg(t),其中μ、ν為常數(shù)。
引理3:[16]假定x(t)∈為連續(xù)可微的向量函數(shù),則對任意t≥t0,存在對稱正定方陣P使得下式成立:
定理1:若存在對稱正定方陣P,正對角矩陣Γ=diag[γ1,γ2,…,γn]>0,使得下面LMI成立:
則系統(tǒng)是全局漸近穩(wěn)定的。
顯然F(0)=0。因此,為研究系統(tǒng)(1)的穩(wěn)定性,只需研究系統(tǒng)(5)零解的穩(wěn)定性。
下面構(gòu)造函數(shù)
其中矩陣P滿足引理3,則
進一步,
為系統(tǒng)(1)選取參數(shù)
通過Matlab軟件LMI工具包進行求解,結(jié)果顯示LMI條件是可行的,可行解如下:
根據(jù)定理1可知系統(tǒng)是全局漸近穩(wěn)定的,圖1~3分別為x1(t)、x2(t)、x3(t)隨時間t(s)收斂于0的趨勢圖。
圖1 x1(t)隨時間t收斂于0的趨勢圖Fig.1 Tendency chart ofx1(t)converging to 0 witht
圖2 x2(t)隨時間t收斂于零的趨勢圖Fig.2 Tendency chart ofx2(t)converging to 0 witht
圖3 x3(t)隨時間t收斂于零的趨勢圖Fig.3 Tendency chart ofx3(t)converging to 0 witht
本文通過對分?jǐn)?shù)階Hopfield型神經(jīng)網(wǎng)絡(luò)進行全局漸近穩(wěn)定性分析,給出了判定該類系統(tǒng)全局漸近穩(wěn)定性的LMI條件,并進行了Matlab仿真,通過實例驗證了文中定理的正確性。
[1]PODLUBNY I.Fractional differential equations:an introduction to fractional derivatives,fractional differential equations,to methods of their solution and some of their application[M].New York:Academic Press,1998:78-81.
[2]KILBASAA,SRIVASTAVAH M,TRUJULLO J J.Theory and applications of fractional differential equations [M].Amsterdam:Elsevier Science Limited,2006:90-99.
[3]LAKSHMIKANTHAM V,VATSALA A S.Basic theory of fractional differential equations[J].Nonlinear Analysis:Theory,Methods andApplications,2008,69:2677-2682.
[4]LAKSHMIKANTHAM V.Theory of fractional functional differential equations[J].Nonlinear Analysis:Theory,Methods andApplications,2008,69:3337-3343.
[5]AGARWAL R P,LAKSHMIKANTHAM V,NIETO J J. On the concept of solution for fractional differential equations with uncertainty[J].Nonlinear Analysis:Theory,Methods andApplications,2010,72:2859-2862.
[6]LI C,CHEN A,YE J.Numerical approaches to fractional calculus and fractional ordinary differential equation[J]. Journal of Computational Physics,2011,230:3352-3368. [7]MOROGADO M L,F(xiàn)ORD N J,LIMA P M.Analysis and numerical methods for fractional differential equationswith delay[J].Journal of Computational and Applied Mathematics,2013,252:159-168.
[8]ZHAO Y G,WANG Y Z,LIU Z B.Finite time stability analysis for nonlinear fractional order differential systems [C]//The 32ndChinese Control Conference.Xi’an:IEEE CPP,2013,487-492.
[9]LIAO Z,PENG C,LI W,WANG Y.Robust stability analysis for a class of fractional order systems with uncertain parameters[J].Journal of the Franklin Institute,2011,348:1101-1113.
[10]LI Y,CHEN Y,PODLUBNY I.Mittag-leffler stability of fractional order nonlinear dynamic systems[J].Automatica,2009,45:1965-1969.
[11]LI Y,CHEN Y,PODLUBNY I.Stability of fractional-order nonlinear dynamic systems:Lyapunov direct method and generalized Mittag-Leffler stability[J].Computers and Mathematics withApplications,2010,59:1810-1821.
[12]BOROOMAND A,MENHAJ M B.Fractional-order hopfield neural networks[J].Advances in Neuro-Information Processing,2009,5506:883-890.
[13]KASLIK E,SIVASUNDARAM S.Nonlinear dynamics and chaos in fractional-order neural networks[J].Neural Networks,2012,32:245-256.
[14]CHEN L,CHAI Y,WU R,et al.Dynamic analysis of a class of fractional-order neural networks with delay[J]. Neurocomputing,2013,111:190-194.
[15]ZOU T,QU J,CHEN L,et al.Stability analysis of a class of fractional-order neural networks[J].Telkomnika Indonesian Journal of Electrical Engineering,2014,12:1086-1093.
[16]MANUEL A DUARTE-MERMOUD,N AGUILA-CAMACHO,JAVIER A GALLEGOS,et al.Using general quadratic Lyapunov functions to prove Lyapunov uniform stability for fractional order systems[J].Commun Nonlinear Science Numer Simulat,2015,22:650-659.
Global Asymptotically Stability of Fractional-Order Hopfield Neural Networks
CUI Shiweia,GAI Mingjiub,LIU Xiaoleib
(Naval Aeronautical and Astronautical University a.Graduate student’Brigade; b.Department of Basic Sciences,Yantai Shandong 264001,China)
In this paper,the global asymptotically stability of fractional-order hopfield neural networks was investigated, and a sufficient condition was given by using LMI approach.At last,a numerical example and corresponding numerical simulation were presented to demonstrate the effectiveness of the result.
fractional order;neural networks;asymptotically stability;LMI
O175.13
A
1673-1522(2015)05-0493-04
10.7682/j.issn.1673-1522.2015.05.019
2015-05-10;
2015-07-20
崔世維(1990-),男,碩士生。