Question

关于MATLAB中bp神经网络的问题

我编写了一个程序，是一个三层bp神经网络。可是预测结果不好。下面是程序，请高手解答下。
year=1999:2008;
p=[0.0900 0.5200 -0.3000 -0.0116 0.3956 -0.3298 0.1727 -1.8408 0.9679 0.8843 0.9533 -0.9402 -0.0332 0.8697;...]
0.0800 0.8800 -0.0333 -0.0102 0.4833 -0.0999 -0.0005 -1.5508 0.9412 0.9609 0.8544 0.7759 -0.0625 0.3039;...
0.0760 0.9400 -0.6222 -0.0102 0.5279 -0.0490 0.0082 -1.5241 0.9372 0.9197 0.7588 -0.4692 -0.0671 0.7168;...
0.0990 0.7000 -0.5333 -0.0102 0.5603 -0.0822 -0.1054 -1.3844 0.9408 0.9267 0.7139 -0.4692 -0.0629 0.6669;...
0.0880 0.9600 -0.8444 -0.0101 0.6246 -0.1072 -0.0802 -1.1748 0.9353 0.9323 0.8459 -0.1274 -0.0691 0.5544;...
0.0620 0.4200 -0.3000 -0.0101 0.6138 -0.1230 -0.1932 -1.2061 0.9459 0.9375 0.8285 0.2796 -0.0572 0.5691;...
0.0860 0.8400 -0.6222 -0.0103 0.5779 -0.0357 -0.2561 -1.8357 0.9368 0.9370 0.7468 -0.2528 -0.0674 0.6117;...
0.0960 0.9000 -0.4111 -0.0102 0.5749 0.0798 -0.2294 -1.3910 0.9300 0.7198 0.5827 0.0801 -0.0753 0.4355;...
0.0820 0.2400 -0.5222 -0.0103 0.4494 0.0311 -0.2839 -1.6046 0.9754 0.7503 0.9456 0.3715 -0.0252 0.3791]';
t=[0 0 0 1 0;1 0 0 0 0;0 0 0 0 1;0 0 0 1 0;0 0 0 1 0;0 1 0 0 0;0 0 0 1 0;0 0 1 0 0;0 1 0 0 0]';
[p,minp,maxp,t,mint,maxt] = premnmx(p,t);
normInput=p;normTarget=t;
NodeNum1 = 14;NodeNum2=7;TypeNum = 5;

TF1 = 'tansig';TF2 = 'tansig'; TF3 = 'tansig';
net=newff(minmax(normInput),[NodeNum1,NodeNum2,TypeNum],{TF1 TF2 TF3},'traingdx');
net.trainParam.epochs=10000;net.trainParam.goal=1e-6;PL.lr=0.01;
net.trainFcn = 'traingdm';
[net,tr] = train(net,normInput,normTarget,[],[]);

预测数据是p1=[0.0310 0.0200 -0.6444 -0.0103 0.8563 0.5906 -0.2513 -1.2564 0.9839 0.3357 1.0108 0.6734 -0.0164 0.3014]';
理想预测结果是t1=[1 -1 -1 -1 -1]';当然是误差范围内就行。我邮箱是wcn517@sina.com，谢谢了~

Answer 1

net.trainFcn = 'traingdm'; 把这个去掉就好了，我试了几次可以的！