幾類時滯差分、微分方程神經(jīng)網(wǎng)絡(luò)模型的動力學(xué)分析.pdf

1、學(xué)校代號學(xué)號密級10532B03061002湖南大學(xué)博士學(xué)位論文幾類時滯差分、微分方程神經(jīng)網(wǎng)絡(luò)模型的動力學(xué)分析Y893218堂焦史遣厶壁名!：生塞塹曼堅絲魚墅盟鲞；夔塞室墼捶塑L羞望焦：鏨堂蘭鹽量絲跡堂墮一童些名整!廑旦塑堂——迨塞墁套旦塑：：!!!!壘堡旦墨!旦迨塞釜塑旦塑!!墮生墮旦!!旦簧整重量璽圭盎：旦亟王墊握些耋翌堂莖坌：壁坌互翌塑絲豎竺堡型竺墊壟蘭坌塹AbstractThedynamicsofneuralnetwork，wh

2、ichhasbeenappliedwidely，isthefoundationofapplicationanddesignConsideringthatthereistimelaginthetransferenceofinformationbetweenneuronsintheneuralnetwork，theneuralnetworkmodelsdescribedbydifferenceequationsanddifferential

3、equationsshouldcontaindelaysDuetotheshortageofefficienttoolandmethodinthequalitativeanalysisoflmgedelaydifferenceanddifferentialequations，whilethedynamicalstudyofsmalldelayneuralnetworkmodelscanprovideusefulnmthodandtool

4、toinvestigationoflargenetwork，SOitismeaningfulthatwestudythedynamicsofsmalldifferenceequationsanddifferentialequationsneuralnetworkwithdelaysThisdissertationmainlyexploresthedynamicsofsolutionsofseveraldelaysmallneuralne

5、tworkmodelsdescribedbydelaydifferentialequationsanddelaydifferenceequations，in—cludingtheconvergence，periodieityandstabilityofequilibriumsolutionsofthemodelsThisdissertationisconsistsoffourchapters，Astheintroduction，inCh

6、apter1，thehistorybackgroundofneuralnetworksandmainreserchcontentsofthedissertationarebrieflyaddressedInChapter2，thedissertationdiscussestheconvergenceofsolutionsandtheexistenceofasymptoticallystableperiodicsohitionsofacl

7、assofdelaydifferenceequationswithpieeewiseconstantnonlinearitybymeansoftheinequalityskills，theiterationofaonedimensionalmapandfixedpointtheoremInagivenspaceofinitialfunction，theconvergenceofsolutionofthemodelisprovedfors

8、omedifferentscopesofsignalfunctionthreshold；aexistenceconditionwhichCallguaranteetheexistenceofaasymptoticallystableperiodicsolutionsofthemodelisobtainedInChapter3thedissertationiulprovesadiscrete!timedifferenceneuralnet

9、—workmodeloftwoneurons，andintroducestwodifferentnonlineardiscontimlOllSsignalfunctionswhichdisplayobviouslymeaningInagivenspaceofinitialfunc～tion，theconvergenceofsolutionoftheimprovedmodelisobtainedforbigandcriticalthres

10、holdofsignalfunctionsrespectively；forsmallthreshold，aexistencetheoremofaasymptoticallystablesynchronizedperiodicsolutionsisprovedbyrneallsoftheiterationofaone!dimensionalmapandfixedpointtheoremInChapter4，thedissertationi