1
$\begingroup$

When creating different hyperparameter combinations, does the function evaluate combination 1 on the same fold as combination 2? As in, are the folds the same across combinations? I understand that for example for cv = 3, a third of the dataset will be held out for evaluation. There will be three such splits, each containing a test fold. Will the split be different for another combination? I also think that it is unnecessary computation to split again but does it happen?

I could not find strong reference to this fact in the documentation except for a part about cv_results_ or it is just assumed that, this is the way it is.

$\endgroup$

1 Answer 1

1
$\begingroup$

The probability that there's a mistake in SK-Learn is very low.

Reasons:

  • It's a mature library

  • Among the contributors is Sebastian Raschka. Those guys are very meticulous.

https://sebastianraschka.com/faq/docs/evaluate-a-model.html

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.