As I am still on the journey to understand what when and how to use, I am now at the point how to proceed after finding the best hyperparameters:

  1. Define Model (NN)
  2. Split Data into Train and Test
  3. Hyperparameter optimization by using Train
  4. Split Train into Train2 and Val
  5. Use best estimator and .fit() again on Train2 while validating on Val

Is this correct? The result is then only a single value by checking the loss. In 5. there is nothing changed anymore (no hyperparameters,..) and thus even a learning curve is not possible anymore. Is this procedure right? Do I miss something? It feels a bit not having a single plot after all this modelling and instead only a single number (or two for train and val).


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.