I have a set of data to train with comprising of 100 elements, 50 of which support theory A, and 50 support theory B. It turns out that both theories are valid, but the signal that A is valid is the inverse of the signal that B is valid. There are other possible signals that are invalid so this isn't a case of everything being valid.

As a complete dataset, how will this get treated? Will the two opposite signals resolve into two separate theories, or will they cancel each other out and no theory will be found at all?

Do I need to split the same dataset into random subsets, several times, and train on those as well to reveal the two separate theories? Is this a standard practice used to discover signals that are otherwise lost/cancelled out?

  • $\begingroup$ Hi. Are you trying to solve a binary classification problem? If yes, it doesn't really matter, I think, whether all instances are "valid" (in some sense), provided that they are different in some way, and that's why people labelled them with "A" or "B". $\endgroup$
    – nbro
    Feb 3 at 17:16

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.