時(shí)滯遞歸神經(jīng)網(wǎng)絡(luò)的漸近性與周期性研究.pdf

1、聊城大學(xué)碩士學(xué)位論文時(shí)滯遞歸神經(jīng)網(wǎng)絡(luò)的漸近性與周期性研究姓名：陳紅軍申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：系統(tǒng)理論指導(dǎo)教師：王林山20060401聊城大學(xué)碩士學(xué)位論文AbstractArtificialNeuralNetworks(ANN)hasfounditspromisingpotentialforapplicationsinclassification，parallelcomputing，andassociativememoryespecial

2、lyinsolvingsomeoptimizationproblemSuchapplicationsrelyontheexistenceofequilibriumpointsorofauniqueequilibriumpointandthequalitativepropertiesofstabilityThus，thequalitativeanalysisofthesedynamicbehaviorsisaprerequisiteste

3、pforthepracticaldesignandapplicationofneuralnetworksInthispapertheasymptoticbehaviorsofdelayedrecurrentneuralnetworksarestudiedThepaperincludes6VechaptersThefirstchapterispreliminariesWeintroducethebasicconceptsaboutArti

4、ficialNeuralNetworks，aswellasDynamicalSystemInthesecondchapteraclassofstaticneuralnetworkswithdelaysisproposedByusingthetheoryoftopologicaldegree，M—matrixandLyapunovfunctionalmethods，wediscussitsglobalasymptoticalstabili

5、tyandsomesufficientconditionsensuringexistence，uniquenessandglobalasymptoticaistabilityoftheequilibriumpointarepresentedInthethirdchaptertheglobalexponentialstabilityofaclassofcellularneuralnetworkswithtimevaryingdelaysi

6、sinvestigatedbyusingtopologicaldegree，MmatrixandHalanayinequalitySomesufficientconditionsfortheexistenceandglobalexponentialstabilityoftheequilibriumareobtainedTheresultspresentedherearebetterthanthosegiveninthequotedlit

7、eratureInchapterfourtheglobalexponentialrobustperiodicityisdiscussedforaclassofintervaldelayedneuralnetworksSomesufficientconditionsensuringexistenceanduniquenessofglobalexponentialrobustperiodicsolutionareobtainedbyusin

8、ginequalityanalysistechniqueandLyapunovfunctionalmethodThesecriteriaextendorimprovethepreviouslyknownresultsInthelastchapterwediscussaclassofCohenGrossbergneuralnetworkswithinfinitedistributeddelaysByintroducingtwoHewtyp