搜索
227Vol122,No1720057ComputerApplicationsandSoftwareJul12005:2004-04-29,,:,,,,(510320),,,,,,,,AKINDOFACCORDINGTOTHEDESIGNOFTHECOMPETITIVELEARNINGMODELBASEDONWAVELETTYPEFUNCTIONAL2LINKNEURALNETWORKANDITSAPPLICATIONMaoGuangxi(DepartmentofInformation,GuangdongBusinessCollege,GuangzhouGuangdong510320,China)AbstractIntheneuralnetworkmodeclassificationmethod,oneoftherelativelyfamousmethodsisthatcompetitivelearning.Competi2tivelearningadoptthehierarchicalstructuresofunits,carryonthecompetitionbyoneakindofwaytargetthatpunishatpresentsystemonvictordrawcompletely.Thecommonneuralmethodofnetworkmodeclassificationmethod,becauseofitsrelativelylongtreatmenttimeandcomplexityofthedata,itisverydifficulttosuitableforthelarge2scaledatabase.Forthisreason,thisliterarygraceatlittlewavebasesthatmodulatetoinputtingmodepretreatment,havedesignedthecompetitionbasedonwavelettypefunctional2linkneuralnetworkandstudythesystemonthebasisoffunctional2linkneuralnetwork,fullyutilizetheadvantageofthesmallwavevaryingandfunctionchainneuralnetwork,systemthatdesignlikethishavesurprisingstudyspeed,systemthecommonabilitycharacteristicfine,betteradaptabilitiesofstructure.KeywordsWaveletFunctional2linkneuralnetworkCompetitivelearning11.1,[6](),1,(),(),,,,,,,,,,11.2[4]:,k,Ok,f1(Ok),f2(Ok),,fn(Ok)Sobajic1988[4],,©1994-2006ChinaAcademicJournalElectronicPublishingHouse.Allrightsreserved.()2.12.1.1,2.1.2,,,Okf1(Ok),f2(Ok),,fn(Ok)WickerhauserCoifmanShannon[5]:HHilbert,vH,v=1,H=(+)HiH,:(v,{Hi})=(-vi2Lnvi2)1/2V{Hi}Shannon,{vi}v{Hi}Shannon,H2.2222.2.1,Coifman,2.2.2,,,,,Shannon2.2.3:,,,(Hebb)2.2.4,,2.2.5(,),/2.3(1),(2)V(),{H(k)i},V,V=V(k)iH(k)i,k,V(k)ii(3){V(k)i},,{V(k0)i},{H(k0)i}(4){H(k0)i(V)},(5),,,,,2.4,,:(1),,,(2),,,(3),,,,(4),,3,,©1994-2006ChinaAcademicJournalElectronicPublishingHouse.Allrightsreserved.(:xi,yi),21xiyixiyi10.24110.3320.18120.5730.46130.4940.46140.4550.66150.0160.24160.0470.75170.1780.12180.2790.66190.03100.71200.1225000.0076920.23076910000.0007690.03076915000.0003850.0230772000X0.0192312500X0.0153853000X0.0076923500X0.0038464000XX,,,,,,,,,[1]MehmedKantardzic,DATAMININGConcepts,Medels,Methods,andAlgorithms.,20038.[2]JiaweiHan,MichelineKamber,DataMiningConceptsandTechniques.,20018.[3]CoifmanR.WicherhauserM.V.,Eutropy2basedalgorithmsforbestbasisselection.1992,38IEEETrans.27,713718.[4]SobajicD.,NeuralNetsforControlofPowerSystems,Ph.D.Thesis,ComputerScienceDept,CaseWesternKeserveUniversity,Cleveland,1988.[5]KlassenM.S.,PaoY.H.,Characteristicsofthefunctional2linknet:Ahigherorderdeltarulenet.IEEEProc.of2ndAnnualInternationalCon2ferenceonNeuralNetworks,1988,I,507513.[6]D.E.RumelhartandD.Zipser,Featurediscoverybycompetitivelearn2ing.CognitiveScience,9:75112,1985.(13)(x,y),():(x,y)+(1/4m,0)(x,y)+(1/4m,1/4m)(x,y)+(0,1/4m)4)|U|15)2.4,,[2],SierpinskiHausdorff,,,,,,(),,,,,/[3],:1)f1f2;2)f1f2c1;3)f1f2c2,,pm=0.0013,,50,100,0.9,0.001:1)m4,100%,02)m=4,33)m=5,134)m=6,,21106,m,,m,,(),,,,,,,[1],[M],.[2],[M],.[3],[M],.[4],SierpinskiHausdorff[J],(A),1999,29,(2):138144.[5],V2[J],,2001,1:13.'1994-2006ChinaAcademicJournalElectronicPublishingHouse.Allrightsreserved.