Why it is diferent the result?

Good afternoon

I have a question about the neuronal network toolbox, I trained a network with 7 inputs and 44 outputs using the method backpropagation (trainlm), the network has 2 layers, and once trained, I compare the result using the funtion Y = SIM(net,P) against the mathematical model (Y = purelin(LW*tansig(b1 + IW*P) + b2)), where b1 and b2 are the biases matrix, IW is the weight matrix of hidden layer, LW is the weight matrix of output layer and P is the input vector, the problem is ¿why the results differ?, should not be the same or approximate?.

Appreciate any kind of response!!

Note: Excuse for my inglish

Answer 1

[Greg Heath]

The answers will be the same, PROVIDED you take into account that the network automatically transforms the data to [-1,1] during training.

help mapminmax

doc mapminmax

You can make life easier by transforming the data yourself before and after training the net.

Another approach is to prevent the automatic scaling via something like (e.g., not exactly)

net.inputs.processFcn = '';

Obviously, using mapminmax before and after training is easier.

Hope this helps.

Thank you for formally accepting my answer.

Greg