基于遺傳算法優(yōu)化神經(jīng)網(wǎng)絡的混沌控制方法.pdf

1、鄭州大學碩士學位論文基于遺傳算法優(yōu)化神經(jīng)網(wǎng)絡的混沌控制方法姓名：姜顯揚申請學位級別：碩士專業(yè)：通信與信息系統(tǒng)指導教師：王忠勇20040501AbstractAbstractTherearegreattheoreticandpracticalvaluesinresearchingchaosandchaoscontrollingAnewkindofmethodsispresentedforcontrollingchaoticdynamica

2、lsystemsusingRadialBasisFunction(RBnneuralnetworksbasedonGeneticAlgorithms(GA’s)，calledGANNlearningmethodWiththegeneralpurposestochasticsearchcapability，GA’Soptimizetheneuralnetwork’Sstructureparameters，andwiththestrongn

3、onlinearapproachingcharacter，theneuralnetworkcanlearntoproduceaseriesofsmallperturbationstoconvertchaoticoscillationsofadynamicalsystemintoaperiodicorbitAnentirelyunsupervisedlearningstrategyisadopteddirectlySOthatthesys

4、temknowledgeisnotneedtoberealizedinadvanceInsomerealworldphysicalchaoticsystems，itisdifficulttodeterminethekeyparameters，SOtheproposedmethodcanbeappliedtomorepracticalsituationsThealgorithmconvergenceperformancehasbeenan

5、alyzedcarefullyandtheproofshavebeengiveilComputersimulationshavebeena／soconductedtocontroltwochaoticsystems，ie，theH6nonmapandthelogisticmapTheresultsindicatedthatthismethodiseffectiveThemaincontentsofthisthesisareasfollo

6、ws：(1)ThemainconceptsofchaosandthekeymethodsofcontrollingchaosarediscussedThenewresearchaboutthisadvancedresearchareaisalsosummarizedbasedonthedemandofourtasks(2)ApplyingGA，whichhasgeneralpurposeoptimizationtrait，tothene

7、uralnetworks’leaming，anewmethodforcontrollingchaoticsystemcalledGANNcontrollingmethodisproposed(3)TheconvergenceoftheGANNcontrollingsystemisanalyzedandprovedSOastodeterminethefeasibilityofthisnewmethodtheoretically(4)Ane

8、wkindofmodifiedadaptiveGAispresentedInthisalgorithm，anewoperatorcalledsimplificationisapplied(5)ApplyingthismodifiedGAtomultimodalfunctionoptimization，thesimulationresultsindicatesthatthismethodisfeasible，andthemodifieda