時(shí)滯神經(jīng)網(wǎng)絡(luò)系統(tǒng)的穩(wěn)定性及狀態(tài)估計(jì)研究

1、大連理工大學(xué)碩士學(xué)位論文時(shí)滯神經(jīng)網(wǎng)絡(luò)系統(tǒng)的穩(wěn)定性及狀態(tài)估計(jì)研究姓名：王紅申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：控制理論與控制工程指導(dǎo)教師：楊斌20081201時(shí)滯神經(jīng)網(wǎng)絡(luò)系統(tǒng)的穩(wěn)定性及狀態(tài)估計(jì)研究ResearchoilStabilityandStateEstimationforNeuralNetworkSystemswithTimeDelayAbstractNeuralnetworks(NNs)haveappliedinavarietyofareas

2、，suchassignalprocessing，patternrecognition，staticimageprocessingassociativememory，andcombinatorialoptimizationinthepastdecadesIntherealworld，timedelayisfrequentlyencounteredinNNs，itcaneasilyreducethetransmissionspeedande

3、vencauseinstabilityandoscillationsinasystemSothestudyontheneuralnetworksystemswithtimedelayisimportantinboththeoryandpractice，andhasthusbeenofgreatinteresttoalargenumberofresearchersinthelastfewdecadesThepurposeofthisdis

4、sertationistostudystabilityproblemandpresentthemorerelaxedstabilityconditionsforneuralnetworkswithtimedelaysInthefirstchapter，thedevelopmentandthemainmethodusedtohandledelayedneuralnetworksareintroducedInthesecondchapter

5、，anewaugmentedLyapunovfunctionalisproposedbyconsideringtheintegralofthecrossproducttermofthestate，adelaydependentcriterionfordelayedneuralnetworksisobtainedbyemployingfree—weightingmatrixapproachAtthesametime，thesystemis

6、furtherstudiedbyusingtheinformationofthedelayfractioningmethodsThenewintegralineqoalitiesareusedtoobtainthemorerelaxedstabilitycriterionforthedelayedneuralnetworksInthethirdchapterthedelayrangedependentstabilityproblemfo

7、rNNsistakenintoaccountTheinformationofthelowerandupperboundsofdelayisusedwhenconstructingtheLyapunovfunctional，anewmethodthatconsiderstherelationshipbetweenthetimevaryingdelayanditslowerandupperboundsisproposedwhenestima

8、tingtheupperboundofthederivativeofLyapunovfunctionalLessconservativedelay—range—dependentstabilitycriteriaforNNswithtimevaryingintervaldelayarepresentedThecriteriaarealsoextendedtosystemswithtime—varyingstructureduncerta

9、inties，thelessconservativerobuststabilitycriterionisobtainedInthefourthchapter，byintroducinganewLyapunovfunction，thedelay—dependentstabilityforcontinuoussystemswithtwoadditivetime—varyingdelaycomponentsisstudiedadelaydep

10、endentsufficientconditionwhichcanguaranteethesystemasymptoticallystableisderivedbasedonthetermoflinearmatrixinequality(LMI)approachAnexampleshowsthattheconclusionobtainedinthischapterhaslessconservativenessthantheexistin