search

ray-project / tune-sklearn

A drop-in replacement for Scikit-Learn’s GridSearchCV / RandomizedSearchCV -- but with cutting edge hyperparameter tuning techniques.
https://docs.ray.io/en/master/tune/api_docs/sklearn.html
Apache License 2.0
465 stars 52 forks source link

Implement resource step #121

Closed Yard1 closed 3 years ago

Yard1 commented 3 years ago

This PR changes the behavior of how warm_start sklearn estimators are handled by changing the value by which the resource parameter (max_iter or n_estimators) is incremented from 1 to base_estimator.RESOURCE_PARAM / max_iters, where RESOURCE_PARAM is the resource parameter and max_iters is tune-sklearn's maximum number of early stopping iterations. That way, the differences between each early stopping iteration will become more pronounced, with the last iteration having CV scores equal to real scores.

Yard1 commented 3 years ago

Yeah, those look good to me. Formatter didn't make any changes, but in my experience the linter check is stricter - will wait for CI and fix if needed.

richardliaw commented 3 years ago

thanks for the contribution!!