搜索
:2002201229(710064),,,,,Abstract:Thispaperputsforwardthetrafficflowtimeseriesprognosticationbasedonthestatisticallearningtheory.Comparedwiththetraditionalstatistics,thestatisticslearningtheoryhasagoodclassificationabilityforlimitedtrainingsamples.Thistheoryfeaturesrapidconvergenceandlocalminimumavoidance.Thesimulationexperimentforthetrafficflowatacrossingprovesthevalidityandefficiencyofthismethed.Keywords:statisticallearningtheory;supportvectormachine;timeseriesprognostication;trafficflow0ITS(IntelligentTranspor2tationSystems),[1],,[2],(StatisticalLearningTheory),,,1SVMSVM:,,[3]:1);2);3);4)(),,VC1.1VC,,VC(Vapnik2ChervonenkisDimension)VC:,h2h,h;VCh,VCVC,VC(),VC,VC,()VC[3]1.272,:yx,F(x,y),l:(x1,y1),(x2,y2),+,(xl,yl)(1)XiRN,yi{+1,-1},i=1,+,l{f(x,w)}f(x,w0),:R(W)=L(y,f(x,w))dF(x,y)(2)L(y,f(x,w)),(1),(2),(ERM),:Remp(W)=1l2li=1L(yi,f(xi,w))=1l2li=1ûf(xi,w)-yiû(3),ERM,:l,,ERM1.3(StructuralRiskMinimization,SRM)VapnikChervonenkis:Remp(W)R(W)1-G[3]:R(W)Remp(W)+(h(ln(2löh)+1)-ln(Gö4)l)(4)hVC,l:(),,VC,(4)Remp(W)R(W),,VC,,,Vapnik:,VCRemp(W)1.4SVMSVM,,,,,,[4,5]11.4.1,1,H,H1H2,(margin),,H1H2[5]:Class1Class2,(w,b),(wõxi)+b1,PxiClass1(5a)(wõxi)+b-1,PxiClass2(5b)(w,b),Class1Class2,,yi[(wõxi)+b]1,i=1,+,l,û(wõxi)+bû1,1öûûwûû,ûûwûû2,,ûûwûûA,x1,x2,,xlR,h:822002220(105)hmin{R2A2,N}+1(6),ûûwûû2VC,,,Ni:Ni0,i=1,+,l,(7):yi[(wõxi)+b]1-Ni,i=1,+,l(8),(7)(8)5(w,N)=12(wõw)+C(2li=1Ni)(9)w(C)1,,2(9)Lagrange,LagrangeAi,,:2li=1Aiyi=0(10a)0AiC,i=1,+,l(10b)Ai:W(A)=2li=1Ai-122li,j=1AiAjyiyj(xiõxj)(11),Ai,,,Ai,,:f(x)=2Aiyi(xõxi)+b(12)1.4.2,SVM:5xZ,Z,,Z(x)õZ(y)=K(x,y),,:W(A)=2li=1Ai-122li,j=1AiAjyiyjK(xiõxj)(13)(10),:f(x)=2AiyiK(xõxi)+b(14)K(x,y),,K(x,y)Mercer[6]2,,,,t,y(t-n),y(t-n+1),y(t),y(t+1),:y(t+1)=+(y(t),y(t-1),+,y(t-n)),,[7]n,n+1,1990,,:,,,25,14h:,8:0016:00,:18:0022:00y(t-3),y(t-2),y(t-1),y(t),SVMY,y(t+1)2,16.03à,[2]BP11.18à,292:2001210208;:2001212202(430063)(100080),,,,,,Abstract:BasedOnthetwotypesoffingerprintminutiae,thispaperdesignsaproperpresentmentofcodes,whichmakespossiblethematchingofeachcodeontypes.Thegeneticalgorithmforfingerprintverificationispresented.Computationoftheexamplesshowsthattheveracityandspeedaresatisfactoryfortherealtimefingerprintverification.Keywords:fingerprints;verification;minutiae;geneticalgorithm1,,,:;;;,[13],FBI:,118:0019:0020:0021:0022:0043964163444643944048410239164160415938183,,,,,V.Vapnik,,,[6],,1..,2001(9)2.BP.,2001(3)3..:,20004,..,2000()5VapnikV.Thenatureofstatisticallearningtheory.Springer,19956C.J.C.Burges.Atutorialonsupportvectormachinesforpatternrecognition.KnowledgeDiscoveryandDataMining,19987..,2001(5)032002220(105)