yongzhen913 发表于 2007-10-26 10:01

输出层函数问题

P:%%P 10个时刻为20组热电耦温度数据(10*20)
T;%%根绝10个时刻热电耦的温度进行模式识别(0 到1之间)
=premnmx(P) ;
=premnmx(T);
net40=newff(minmax(P),,{'logsig','logsig','logsig'},'trainlm');
net40.trainParam.epochs=5000;
net40.trainParam.show=1;
net40.trainParam.goal=0.0001;
net40.trainParam.Ir=0.05;
net40=train(net40,Pn,Tn);
TRAINLM, Epoch 0/5000, MSE 0.964332/0.0001, Gradient 3.98924/1e-010
TRAINLM, Epoch 1/5000, MSE 0.856971/0.0001, Gradient 0.0500049/1e-010
TRAINLM, Epoch 2/5000, MSE 0.85521/0.0001, Gradient 1.60191e-082/1e-010
TRAINLM, Minimum gradient reached, performance goal was not met.
最后输出层函数采用'logsig'时老出现以上问题
换成‘purelin’函数时则收敛良好,但预测效果不好?大家帮忙看看呗:)感激阿
页: [1]
查看完整版本: 输出层函数问题