資源描述:
《幾類具變時(shí)滯的神經(jīng)網(wǎng)絡(luò)模型的動(dòng)力學(xué)研究》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。
1、湖南大學(xué)碩士學(xué)位論文幾類具變時(shí)滯的神經(jīng)網(wǎng)絡(luò)模型的動(dòng)力學(xué)研究姓名:彭水軍申請(qǐng)學(xué)位級(jí)別:碩士專業(yè):應(yīng)用數(shù)學(xué)指導(dǎo)教師:黃立宏2002.4.20摘要驢514293本篇論文共由四章組成。第一章概述了問題產(chǎn)生的歷史背景和本文的主要工作。在第二章中,討論一類推廣的具變時(shí)滯和變系數(shù)的雙向聯(lián)想記憶(BAM)神經(jīng)網(wǎng)絡(luò)模型xm)—‘ci(t)xi(t)+aij(t)fj(yy(t))+莖一~(f)))+,’“一,Plj(t)fj(Yj(tIi(t)佰1yj(f)—。dj(t)yj(t)+bji(t)gj(t(f))+∑qⅣ(f)毋(Xi(t—G.ii(f)”+Jj(
2、f),的漸近狀態(tài)。這里,信號(hào)傳輸函數(shù)廠i艚;(f一1,2,?,n,,=L2,?,m)是R—R上的連續(xù)函數(shù),且滿足Lipschitz條件。我們利用M一矩陣?yán)碚?、微分不等式分析技巧和?gòu)造Lyapunov泛函方法,建立了神經(jīng)網(wǎng)絡(luò)系統(tǒng)(E)與時(shí)滯無關(guān)的全局指數(shù)穩(wěn)定性判據(jù)和保證周期解的存在唯一及其穩(wěn)定性的幾個(gè)充分條件,同時(shí)考慮了系統(tǒng)(E)退化為常時(shí)滯和常系數(shù)情形的全局漸近穩(wěn)定性問題,這些結(jié)果較大地改進(jìn)和推廣了一些已知的結(jié)果。第三章研究了具可變時(shí)滯的BAM細(xì)胞神經(jīng)網(wǎng)絡(luò)模型的周期解與穩(wěn)定性問題,獲得了保證系統(tǒng)的平凡解收斂的充分條件以及關(guān)于周期解的存在與穩(wěn)定性
3、的結(jié)論,所得結(jié)果對(duì)于連續(xù)的BAM神經(jīng)網(wǎng)絡(luò)的設(shè)計(jì)和應(yīng)用具有一定的指導(dǎo)意義。關(guān)鍵詞:神經(jīng)網(wǎng)絡(luò),雙向聯(lián)想記憶(BAM),時(shí)滯,平衡點(diǎn),周期解全局漸近穩(wěn)定性,全局指數(shù)穩(wěn)定性,M一矩陣,Lyapunov泛函.AbstractThispaperiscomposedoffourchapters.InChapter1,weintroducethehistoricalbackgroundofproblemswhichwillbeinvestigatedandthemainworkofthispaper.InChapter2,byusingM-matrixtheo
4、ry,someanalysistechniquesandconstructingsuitableLyapunovfunctionals,weinvestigatetheconvergenceofsolutionsandtheexistenceanduniquenessofperiodicsolutionforsystem:P卜_@to卜掣小乃o"+∥m叭”o’∞"“以x(E)舊‘)一dJ(f)YA‘)+善6』f(‘)gj(而(‘))+蚤9F(‘)gj(xi(t-‘Yji(‘)))+Jj(‘),whereand垂arethepropagation
5、alsignalfunctionsdefinedonR.Severalsufficientconditionsguaranteeingtheneuralexponentialstabilityaswellastheexistenceandstabilityofperiodicsolutionoftheneuralnetworkareobtained.Theglobalasymptoticallystabilityofthecorrespondingneuralnetworkmodelswithconstantdelaysandcoefficie
6、ntsisstudiedtoo.Ourresultsimproveandgeneralizesomeknownresults.InChapter3,theperiodicoscillatorysolutionsandtheglobalstabilityarestudiedforaclassofcontinuousbi—directionalassociativememoryneuralnetworkmodelswithvariabledelays,andsomesimpleandnewsufficientconditionsaregivenen
7、suringglobalexponentialstabilityandtheexistenceofperiodicsolutionsoftheneuralnework.TheseresultshaveimportantleadingsignificanceinthedesignandapplicationsofglobalexponentialstableBAMnetworksandperiodicoscillatoryKeywords:neuralnetworks;bi—directionalassociativememory(BAM);de
8、laysequilibrium;periodicsolution,globalasymptoticallystability;globalexpone