annzuoye.m

建立一个BP网络

a=[ 90	25	30	0;
   100	40	50	0;
   200	60	70	20;
   300	120	60	10;
   400	77	31	0;
   500	100	50	10;
   800	280	140	70;
   900	220	60	20;
   1000	540	180	165;
   1200	300	60	40;
   1400	340	100	60;
   1600	660	100	220;
   1800	900	180	180;
   2000	500	100	60]'

t=[0.01 0.01 0.01 0.01 0.99 0.99 0.01 0.99 0.01 0.99 0.99 0.01 0.01 0.99];
%输入输出原值
b=zeros(1,size(a,2));
for i=1:size(a,1)
    b=b+a(i,:);
end
for j=1:size(b,2)
   b(j)=b(j)/size(a,1);
end
%b 为平均值   
for i=1:size(a,1)
   for j=1:size(a,2)
           a(i,j)= (a(i,j)-b(j))/b(j);
           %收入处理 归一化（0，1）之间 
    end
end

net=newff(minmax(a),[6,1],{'logsig','purelin'},'traingdx');
net.trainParam.show=200;%指定算法每个执行表达的训练次数
net.trainParam.lr=0.01;
net.trainParam.epochs=30000;
net.trainParam.goal=0.00001;
net.trainParam.mc=0.59;
net=train(net,a,t);

PP =[600	120	40	20;
     700	160	60	20]' ;
ppp=zeros(1,size(PP,2));
for i=1:size(PP,1)
    ppp=ppp+PP(i,:);
end
for j=1:size(ppp,2)
   ppp(j)=ppp(j)/size(PP,1);
end
%b 为平均值   
for i=1:size(PP,1)
   for j=1:size(PP,2)
        
           PP(i,j)= (PP(i,j)-ppp(j))/ppp(j);
           %收入处理
        end
end
y=sim(net,PP)
output = y

weights11 = net.iw{1,1}
weights = net.lw{2,1}
threshold = net.b{1}
threshold = net.b{2}