Using Optuna hyperparameter importance on Ray Tune experiment with Optuna as search algorithm

Ray Libraries (Data, Train, Tune, Serve) Ray Tune
1

How severe does this issue affect your experience of using Ray?

  • None: Just asking a question out of curiosity

Optuna has a handy function [optuna.importance.get_param_importances](https://optuna.readthedocs.io/en/v1.3.0/reference/importance.html) that can be used on an optuna.study object.

I was wondering how one may go about using this function on a Ray Tune experiment where the search algorithm is optuna, so that I can get which parameters were considered important by optuna?

2

Hey @stephano41, ah I don’t think we have something like this for Ray Tune.

The optuna study is saved as a private attribute in the OptunaSearcher, so a hacky way to do this would be to pass in optuna_search._ot_study into get_param_importances.

But let me create a Github issue to track this feature. [Tune] Support for Optuna `get_param_importances` or equivalent · Issue #25522 · ray-project/ray · GitHub.

1 Like