suanfa.rar_matlab 神经网络

1． 随机逼近算法模块 function noisetest() n=1:64; specimena=sin(n/64*2*pi); specimenb=sawtooth((n+16)/64*2*pi,0.5); w(1:64)=0.05; error=1; k=1; step=0.01; tag=1; while(error^2)>0.001 if(tag==1) y=w*specimena.'; d=1; error=d-y; w=w+2*step*error*specimena; end if(tag==0) y=w*specimenb.'; d=0; error=d-y; w=w+2*step*error*specimenb; end e(k)=error^2; tag=~tag; k=k+1; end plot(e) grid 2． 抗噪检验模块 w t=0.0025; k=1; while k<=15 g=normrnd(0,t*k,[1,64]); specimena1=specimena+g; specimenb1=specimenb+g; ya1=w*specimena1.'; yb1=w*specimenb1.'; error1=(abs(1-ya1)+abs(0-yb1))/2; errorn(k)=error1^2; k=k+1; end i=1:15; plot(i*t,errorn) grid 3．最陡下降法模块 clear; n=1:64; specimena=sin(n/64*2*pi); specimenb=sawtooth((n+16)/64*2*pi,0.5) w(1:64)=0.05; step=0.015; k=1; error=1; while error>0.0001 ya=w*specimena.'; errora=1-ya; producta=errora*specimena; yb=w*specimenb.'; errorb=0-yb; productb=errorb*specimenb; w=w+2*step*(producta+productb); error=errora^2+errorb^2; J(k)=error; k=k+1; end plot(J) grid