Just to check if everything is working properly in my neural net, I set my validation data to be the same as my training data, expecting to achieve a better NRMS for validation data (since it uses only the most recently updated weights for calculating the NRMS). But that didn't happen. More often than not, the NRMS was larger for the validation data than for training data. Am I missing something or is this a normal thing? Cause it doesn't make sense to me.

  • $\begingroup$ Are you sure that your validation set is exactly equal to your training dataset and that you're computing the same metric to measure the performance on both sets? If yes, are you using the same batch size in both cases? Is there any stochasticity in your neural network? $\endgroup$
    – nbro
    Aug 8 '20 at 13:05
  • $\begingroup$ Yes, data is the same and batch sizes are the same. There is no stochasticity. $\endgroup$ Aug 8 '20 at 15:01
  • $\begingroup$ You're doing something wrong then. There's a bug in your code somewhere or you're not doing what you think you're doing. $\endgroup$
    – nbro
    Aug 8 '20 at 15:58

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.