normalizing for neural net fitting?

i have 405 data (value) that i normalized them with matlab function or (formula) and i gave it to neural net fitting to train it and i got an output...the qustion is how do i unnormalize the (output) ?

avg1=(sum(alltargets(1,:))) / 405;


S1= std(alltargets(1,:));


for i=1:405
    Z(1,i)=((alltargets(1,i)-avg1)/S1);
end

Answer 1

[Drew]

It looks like your question is under-specified. What is the design of the neural network? What are the outputs of the neural network designed to represent? For example, is the neural network being used for classification, regression, feature extraction, or something else?

Another idea: Depending on what task you are trying to accomplish, you might gain guidance for your workflow by using an app like the Classification Learner or Regression Learner from the SMLT toolbox, or an app from the Deep Learning toolbox.

[arian hoseini]

i don't know much about neural network but let me explain m work for u...i have an input that contains 5 type of data (1.humdity/2.SDD/3....) that each of them has 405 value...and i have a target that contains 3 harmonic(first harmonic third and fifth) that has 405 value too....i normalize input and target and give them to neural net fitting then i get an output that i need to unnormalize it then gain the FFT from that and compare it wiith other FFT that i have.