optuna.terminator.RegretBoundEvaluator

class optuna.terminator.RegretBoundEvaluator(top_trials_ratio=0.5, min_n_trials=20, seed=None)[source]

An error evaluator for upper bound on the regret with high-probability confidence.

This evaluator evaluates the regret of current best solution, which defined as the difference between the objective value of the best solution and of the global optimum. To be specific, this evaluator calculates the upper bound on the regret based on the fact that empirical estimator of the objective function is bounded by lower and upper confidence bounds with high probability under the Gaussian process model assumption.

Parameters:

gp – A Gaussian process model on which evaluation base. If not specified, the default Gaussian process model is used.
top_trials_ratio (float) – A ratio of top trials to be considered when estimating the regret. Default to 0.5.
min_n_trials (int) – A minimum number of complete trials to estimate the regret. Default to 20.
min_lcb_n_additional_samples – A minimum number of additional samples to estimate the lower confidence bound. Default to 2000.
seed (int | None)

For further information about this evaluator, please refer to the following paper:

Automatic Termination for Hyperparameter Optimization

Note

Added in v3.2.0 as an experimental feature. The interface may change in newer versions without prior notice. See https://github.com/optuna/optuna/releases/tag/v3.2.0.

Methods

evaluate(trials, study_direction)