Regressin equation from neural network does not match to net(x)

Hello,

I have created the codes using the button in the nnstart toolbox 'simple script' as below. After training I have compared the values using following commands and it was showing me different results.

• y1 = b2 + LW*tanh(b1+IW*x)
• y2 = net(x)

Could someone please help me what the promblem is?

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

load('Input_Output.mat') % Data attached

inputs = Input_transpose; targets = Output_transpose;

hiddenLayerSize = 10; net = fitnet(hiddenLayerSize);

net.divideParam.trainRatio = 70/100; net.divideParam.valRatio = 15/100; net.divideParam.testRatio = 15/100;

[net,tr] = train(net,inputs,targets);

outputs = net(inputs); errors = gsubtract(targets,outputs); performance = perform(net,targets,outputs)

view(net)

 b1 = net.b{1};
 b2 = net.b{2};
 IW = net.IW{1,1};
 LW = net.LW{2,1};
 x = Input_transpose(:,1)
 y = b2 + LW*tanh(b1+IW*x)
 net(x)

[Mohamed El Ibrahimi]

https://fr.mathworks.com/matlabcentral/answers/311783-neural-network-how-can-i-get-the-correct-output-answer-without-using-the-function-sim-neural-net#answer_242993

Answer 1

[Greg Heath]

You did not take into account the default minmax normalization of inputs and targets

See

http://www.mathworks.com/matlabcentral/answers/80928-expression-of-ann-model

Hope this helps.

Greg

[Keunwook Chung]

Thank you very much!