Does Matlab make an automatic normalization for the input of a network?

Hello everyone,

I am using the function feedforwardnet for my problem. The values are in the inputs very small (7.14078e-09) and I am asking myself if Matlab makes automatic normalization for these values, or should I do it by oneself?

If Matlab does this automatically, could you please refer that, where I find it in the documentation.

Thank you in advance

Answer 1

[Sai Bhargav Avula]

Most of the neural network creation functions in the toolbox, like feedforwardnet, automatically assign processing functions to the network input and outputs. These are called process functions and can be overridden after the network.

For this case, the feedforwardnet has removeconstantrows and mapminmax as the processing functions. The mapminmax does normalize the input in the range [-1 1] for your data.

For more details refer the following link for input and output process functions.

https://www.mathworks.com/help/deeplearning/ug/choose-neural-network-input-output-processing-functions.html

[Osama Tabbakh]

Thanks.

Could you please take a look for my other question, maybe you have also an answer for me?

https://de.mathworks.com/matlabcentral/answers/469409-why-sets-matlab-automatically-the-activation-functions-for-a-neural-network-like-this?s_tid=prof_contriblnk

[Benze Wang]

The situation I have is that i need to normalize the data in our own way. So is there anyway to cancel the automatic normalization function of feedforwardnet so that I can creat our own?