It occurs to me as I'm normalizing the features for my CNN that I am inadvertently taking the "future" into account by normalizing using the min and max of the entire time period. In other words, at any given point in time, the min and max may be different than what the eventual min and max of the entire time period was, yet the normalized feature will have the "future" values of min and max baked into it after normalization. Is this okay? Should I do a rolling normalization using only the most recently "known" min and max?

  • $\begingroup$ Your question is lacking a lot of important details. What are you trying to model? What is you training and test set split? $\endgroup$ – Alex R. Mar 21 at 18:31

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.