optuna-fast-fanova 0.0.4

optuna-fast-fanova

optuna-fast-fanova provides Cython-accelerated version of FanovaImportanceEvaluator.

The benchmark script was run on my laptop (Macbook M1 Pro) so the times should not be taken precisely.

Installation

Supported Python versions are 3.7 or later.

$ pip install optuna-fast-fanova

Please note that this library depends on the scikit-learn's C-API (Cython pxd files). However, its ABI may contain breaking changes, even in patch releases. If you install optuna-fast-fanova with scikit-learn v1.1.1 and then upgrade scikit-learn to v1.1.2, optuna-fast-fanova will not work. Please reinstall optuna-fast-fanova if you update scikit-learn.

Usage

Usage is like this:

import optuna
from optuna_fast_fanova import FanovaImportanceEvaluator


def objective(trial):
    x = trial.suggest_float("x", -10, 10)
    y = trial.suggest_int("y", -10, 10)
    return x ** 2 + y


if __name__ == "__main__":
    study = optuna.create_study()
    study.optimize(objective, n_trials=1000)

    importance = optuna.importance.get_param_importances(
        study, evaluator=FanovaImportanceEvaluator()
    )
    print(importance)

You can use optuna-fast-fanova in only two steps.

1. Add an import statement: from optuna_fast_fanova import FanovaImportanceEvaluator.
2. Pass a FanovaImportanceEvaluator() object to an evaluator argument of get_param_importances() function.

How to cite fANOVA

This is a derived work of https://github.com/automl/fanova. For how to cite the original work, please refer to https://automl.github.io/fanova/cite.html.