If I am training a neural network and have normalized the data before-hand, should I reverse the normalization to calculate the loss?

This tutorial provides an example of this method.

What if I'm using batch normalization? From this post it looks like there are ways to reverse it but my question is more along the lines of should I bother.

My first thought was that it would depend on the problem but I could not think of a problem where I would not want to minimize the error in the actual units of the problem

  • $\begingroup$ Am not sure about error-term. Are you talking about sampling error or measurement error or outlier $\endgroup$ – Subhash C. Davar May 23 '20 at 7:51
  • $\begingroup$ I mean the loos function fed to model.compile(). Just corrected it in the post $\endgroup$ – seeiespi May 26 '20 at 3:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.