神经网络课件神经网络课51教案n.ppt

文档介绍：,DiscreteHNN(DHNN)differentialequationContinuousHNN(CHNN)Applications:,1N1N2Nn-1Nnw2n…x1x2xn-1xnwn-1,nwn-1,1wn,2wn-1,2……v1(t)v2(t)vn(t)y1y2yn-z-1z-1z-1w11w1nwn1wnn1---2nu2(t)u1(t)un(t)x1(t+1)x2(t+1)xn(t+1)x1(t)x2(t)xn(t)……Computingformula:(series)patternxi(t+1)=sgn[ui(t)]xj(t+1)=xi(t)(ji)(parallel)patternxi(t+1)=sgn[ui(t)](i=1,2,…,n)“energyfunction”Matrixformat:x=1or–1,wandarelimitedconstant,E=E(t+1)-E(t):seriespattern,wij=wji,wkk0E=E(t+1)-E(t)xk=xk(t+1)-xk(t)onlykthnodeisenabledxk(•)=1wij=wji∵∵wkk0E0Theenergyfunctionislimitedconstant,workcanconvergetoalocalminimumpoint.∵-associationMemoryTheMsamples({xi},i=1,2,…,M)’=x+v,xisoneoftheMsamplesvisbiasitemThenoutputY=x-associationMemoryTherelationshipbetweentwosamples:Xiyi,i=1,2,…,MForexample,xistandforaman’spicture,’=x+vthenoutputY=putingphases:Learning:formingW(correspondtox1,x2,…,xp)Initial:v(0)=x(assuminginputx=xk)Running:v(t+1)=sgn(v(t)W)Steadyoutput:v=xk