Neural network Help: Urgent

1 次查看（过去 30 天）

显示更早的评论

Ahmed 2012-1-22

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/26752-neural-network-help-urgent

在 MATLAB Online 中打开

Dear Colleagues

Hope my email finds you well. I made a code using m-file in Matlab to obtain the optimum number of hidden neuron. So that I made a loop to try the performance (MSE) from the number of inputs to the twice number of inputs.

My problem is that the program calculate save the results of last iteration only in work space. So that I need advice about:

1. How to save the results after each iterations in the work space.

2. How to plot the performance graphs after each iteration separately.

3. How to plot the total number of hidden neurons (i) versus MSE in each iterations and let the program calculate the best number of hidden neuron.

[I N]=size(input);
for i=I:I*2
network=newff(input,target,i,{'logsig' 'purelin'},'trainlm');
view(network)
% NN training
network=init(network);
[network,tr]=train(network,input,target);
% Testing the Neural Network
y=sim(network,input);
residual=target-y;
figure
 plot(residual)
 title('Prediction Errors');
 xlabel('Pattern Sequence');
 ylabel('Residuals');  
% NN weight and bias matrices
Layer1_Weights=network.iw{1};
Layer1_Bias=network.b{1};
Layer2_Weights=network.lw{2};
Layer2_Bias=network.b{2};
Layer1_Weights;
Layer1_Bias;
Layer2_Weights;
Layer2_Bias;
end