无界可变延迟随机神经网络的指数稳定性.pdfVIP

应 用 数 学 MATHEMATICA APPLICATA 2014，27(4)：851-857 ExponentialStabilityofStochasticNeural Networkswith UnboundedTime-v‘aryingDelays MENGXuejing(~雪井) (SchoolofStatistics，HubeiUniversityofEconomics，Wuhan430205，China) Abstract：Theexponentialstabilityofstochasticneuralnetworkswithunboundedtime-varyingde laysisinvestigatedwiththehelpofLyapunovfunctionandthesemimartingaleconvergencetheo ram．ThesufficientconditionstOguaranteethealmostsureexponentialstabilityandexponentialsta’ bilitvinmeansquareofatrivialsolutionaregiven．Themethodandresultshereonunboundeddelay systemscovertheusualboundeddelayasaspecialcase． Keywords：Stochasticneuralnetwork；Unboundeddelay；Exponentialstability CLCNumber：O175 AMS(2000)SubjectClassification：60H10；34K2O Documentcode：A ArticleID：1001-9847(2014)04—0851-07 1．Introduction Hopfieldneuralnetworksmodel 士()一一Ax(￡)十Bg(x(t))，t≥ 0 wasfirstly introducedby Hopfieldc in early 1980s．Since then，Hopfield neuralnetworks haveattractedmuch attentiondueto theirpromisingpotentialfortasksofclassification，as— sociativememory，parallelcomputationandtheirabilitytosolvedifficultoptimizationprob— lems．Theseapplicationsarebuiltontheexistenceofequilibrium pointsandtheirstability．In realnervoussystems，stochasticeffects，whichmayleadtoinstabilityofnetworks，shouldbe takenintoaccount[]．In[4]，LIAOandMAOinitiatedthestudyofstabilityandinstability ofstochasticneura1network(SNN)： dx(t)一 [一Ax()+B9(x(t))3dt+g(z())d训()，t≥0． Then，inI-5-1，theycontinuedtheirdiscussionontheSNNswithdelay dx()一 [一Ax()+B (z())]dt+g(z()，z(￡))dw(t)，t≥0． However。theaboveresearchesarerestrictedtoboundeddelaycases．Insomepracticalappli— cationsofneuralnetworks．theinevitabletimedelaymaybeunboundedandvariable引．So，it isimportanttostudythestabilityofSNN withunboundedtime—varyingdelays． * Receiveddate：Dec30，2013 Foundatio