%輸入向量P和目標向量T
P=[0	0	0.62	0	0	0	0;
0.3915	0.4741	0.77	0.5	0.5	1	0.3158;
0.2835	0.5402	0.68	0	0.5	1	0.3158;
0.6210	1.0000	0.63	1	0.5	1	1.0000;
0.4185	0.4183	0.67	0.5	0	1	0.7368;
0.2160	0.4948	0.71	0	0	1	0.2632;
0.9990	0.0383	0.75	0.5	1	1	0.9474
0.5805	0.4925	0.71	0	0	0	0.3684;
0.0810	0.0692	0.76	0	0	0	0.0526;
0.3915	0.1230	0.98	0.5	0	0	0.8974]';
T=[0 0.5313 0.5938 0.9375 0.4375 0.5000 1.0000 0.3750 0.3125 0.6563];
%測試向量P_test和實際輸出T_test
P_test=[0.0270	0.0742	0.62	0	0	0	0.2105;
0.1755	0.3667	0.77	0	0.5	1	0.7368;
0.4320	0.3790	0.68	0.5	0	1	0.2632;
0.4995	0.4347	0.63	0	0	1	0.6842;
0.6885	0.5842	0.67	0.5	0.5	1	0.4211;
0.5400	0.8038	0.71	0.5	0.5	1	0.5789;
0.1620	0.2565	0.75	0	0	1	0.4737]';
T_test=[0.1875 0.4062 0.4375 0.5938 0.6250 0.7187 0.3750];
threshold=[0 1;0 1;0 1;0 1;0 1;0 1;0 1];
a=[10 15 20];
for i=1:3
    net=newff(threshold,[a(i),1],{'tansig','logsig'},'traingdx');
    net.trainParam.epochs =1000;
    net.trainParam.goal=0.001; 
%init函數用于將網絡初始化
    net=init(net);
    net=train(net,P,T);
    Y(i,:)=sim(net,P_test);
end
figure;
%繪制誤差曲線
%中間層神經元個數為10
plot(1:7,Y(1,:)-T_test);
hold on;
%中間層神經元個數為15
plot(1:7,Y(2,:)-T_test,'+');
hold on;
%中間層神經元個數為20
plot(1:7,Y(3,:)-T_test,'--');
hold off;