人工神经网络研究现状及其展望-朱大奇

　第3卷第1期2004年2月　　江南大学学报(自然科学版)JournalofSouthernYangtzeUniversity(NaturalScienceEdition)　　　Vol.3　No.1Feb.　2004　:1671-7147(2004)01-0103-08　　:2003-06-03;　:2003-09-04.:(2002KJ044),(0002133).:(1964-),,,,.,,.人工神经网络研究现状及其展望朱大奇(江南大学通信与控制工程学院,江苏无锡214036)　:回顾了神经网络理论发展的历史和现状。在此基础上,介绍并讨论了20世纪90年代神经网络研究的一些新进展。根据神经网络研究的特点,对人工神经网络今后的发展前景作了一定的评述,人工神经网络的研究与发展必将对现代科学技术产生深远的影响.:人工神经网络;感知器;并行分布处理;模糊神经网络;量子神经网络:TP183:ATheResearchProgressandProspectsofArtificialNeuralNetworksZHUDa-qi(SchoolofCommunicationandControlEngineering,SouthernYangtzeUniversity,Wuxi214036,China)Abstract:Therecentprogressofartificialneuralnetworksisreviewed.Itsnewdevelopmentin90'sisintro-ducedanddiscussedinthispaper.Accordingtothecharacteristicsofartificialneuralnetworks,theauthorpre-sentssomepointsofviewabouttheprospectsofartificialneuralnetworksinthefuture.Theprofoundandfar-reachingeffectcanbepredictedwithdevelopmentofartificialneuralnetworksresearch.Keywords:artificialneuralnetworks;perceptron;paralleleddistributedprocessing;fuzzyneuralnetworks;quantumneuralnetworks　　.400,(Plato)(Asidtole),,.,,,,.2040,、,、,,、,,ANN(ArtificialNeuralNetwork),().,,HechtNielsen,:“,”.、,:[1].,、、、.,,,、、、、、,.1　人工神经网络发展的历史及现状1.1　,3.1.1.1　第一阶段———启蒙时期　.1943,MeCullochPitts,,[2],.MeCullochPitts(,MP),。,Hebb1949Hebb[3].,,.Hebb.1952HodgkinHuxley,H-H.、,.1954,Eccles[4],,.1956,Uttley,.70,,.1958RosenblattMP[5].,,.Rosenblatt,.Rosenblatt,..1960WidrowHoffADALINE[6],,.,;;,Gross-berg,、[7,8];Amari,;Willshaw[9],;Nilsson[10]..1.1.2　第二阶段———低潮时期　,MinskyPapert,1969《Percep-trons》,,,“”.,,,10.,2070,,VonNeumenn,,,.Minsky,,.,,,.1976,GrossbergART(AdaptiveResonanceTheory),.,CarpenterART[11],ART1、ART2ART33ART,ART1104江南大学学报(自然科学版)第3卷　,ART2ART1.ART3-,.1972,.KohonenT,SOM(Self-Organizingfeaturemap);AndersonJ,,“”.KohonenT[12].SOM,、.“”,,,.,.(Ku-nihikoFukushima).,1980“”(Neocognitron).“”,,,.,,,.1.1.3　第三阶段———复兴时期　.1982,Hopfield[13].HopfieldCidUi/dt=∑Ni=1Tijfj[uj]-ui/Ri+Ii,(i=1,2,…,N)　(1)Uii;Ci,Ri;Ii;Tijji;f(u)u.HopfieldLaypunov,Tij=Tji,,.Hopfield,,,,Hopfield[14],,,.,Marr,、,[15].1983,KirkpatrickNP[16],1953Metro-pli.1984,HintonSejnowski,,Boltzmann[17].PiggioMarr,[18,19].Gabor、、、.1986,RumelhartMcCkekkand16《ParallelDistributedProcessing:Ex-plorationintheMicrostructuresofCognition》[20],,,,BP,.,《Perceptrons》,,BP、.PDP,.,MeadConway、Mahowald,,,VLSI.1988,ChuaYang(CNN)[21,22],;Ksko(BAM)[23,24],.2090,,,,.EdelmanDarwinism2090,,DarwinismⅢ,、DarwinNallance,105　第1期朱大奇:人工神经网络研究现状及其展望;1991,Haken,,,,;,,.1994,,[25].,.,,,.Wunsch1990OSAART,.1995,Jenkins(PNN),[26].McAulay[27,28].,.1990Naren-draParthasarathy[29],.(α)[30].Miller(CAMC),[31],,.1993BulsariSigmodial[32],.,.1991Aibara,[33].1991Inoue[34,35]..,,,.,,[36,37].HiroyukiQkada,Sigmoid[38],.、、,,、.CaiYaling3[39],,,,.、、.,,1,.1　Tab.1　ImportantNeuralNetworks(Perceptron)FrankRosenblatt()1958,,,、、(Adaline)BernardWidrow()1960～1962,-、106江南大学学报(自然科学版)第3卷　(Cerellatron)MarrD.()1969～1982,BP(BackPropagation)WerbsP.()RumelhartD.()Mcclelland()1974～1985,,-,,、、ART(AdaptiveResonanceTheory)CarpenterG.GrossbergS.()1976～1990、;.,、BSB(BrainStateinaBox)JamesAnderson()1977,,,,(Neocognition)FukushimaK.()1978～1984,、,SOM(Self-Organizingfeaturemap)TuevoKonhonen()1980,、,HopfieldJohnHopfield()1982,,TSP,(Boltzmanmachine);(Cauc-hyma-chine)HintonJ.()Sejnow-skiT.()1985～1986,;、BAM(Bi-directionalAssocia-tiveMemory)BartKosko()1985～1988,,,CPN(CounterPropaga-tion)RobertHech-tnielsen()1986,,1.2　2080,.19891011—;19902(、、、、、)“”.“,”,300,.,,、.,1987,.(IJCNN).,JournalNeu-ralNetworks,.,.107　第1期朱大奇:人工神经网络研究现状及其展望2　神经网络的发展与展望,、、、、.、.5.2.1　、,,,,.,,,.、、,[40].,.,,.,,.2.2　,,,[42,43].,[41].、、,,,(),.,,,,、.2.3　,,.,.,,,、..,,,,,[48～50];2.4　,.,,.,[42].,.,.,:、、.,,、,.2.5　,,,,.[34,35];,[44～47],.,(Fuzzy)(CMAC),(FCMAC),[51,52],.3　结　语,,.,.,.,108江南大学学报(自然科学版)第3卷　,21.:[1],.[J].,1997,26(5):360-368.[2]MCCLLOCHWS,PITTSW.Alogicalcalculusoftheideasimmanentinnervousactivity[J].BulletinofMathematicalBiophysics,1943,10(5):115-133.[3]HEBBDO.TheOrganizationofBehavior[M].NewYork:Wiley,1949.[4]ECCLESJC.Cholinergicandinhibitorysynapsesinapathwayfrommotor-axoncollateralstomotorneurones[J].JournalPhysiology,1954,126:524.[5]ROSENBLATTTF.Theperception:Aprobabilisticmodelforinformationstorageandorganizationinthebrain[J].PsychologicalReview,1958,65:386-408.[6]WIDROWB,HOFFME.AdaptiveSwitchingCircuits[A].IREWESCONconventionrecord:part4.Computers:MachineSystems[C].LosAngeles:1960,96-104.[7]GROSSBERGS.Ontheseriallearningoflists[J].Bio-science,1969,10(4):201-253.[8]GROSSBERGS.Somenetworksthatcanlearn,rememberandreproduceanynumberofcompialtedspace-timepatternsⅡ[J].StudyAp-pliedMathematics,1970,11(49):135-166.[9]WILLSHAWDJ,BUNEMANOP.Aholographicneuralnetworks[J].Nature,1969,222:960.[10]NILSSONNJ.LearningMachines:FoundationsofTrainablePatternClassifyingSystems[M].NewYork:McGraw-hill,1965.[11]CARPENTERGA,GROSSBERGS.TheARTofadaptivepatternrecognitionbyself-organizingneuralnetwork[J].TransIEEEonCom-puter,1988,15(1):77-88.[12]KOHONENT.Self-organizationandassociativememory[M].NewYork:thirdEditionSpring-Verlag,1989.[13]HOSPFIELDJ.Neuralnetworksandphysicalsystemswithemergentcollectivecomputerab