c-TPE; Tree-structured Parzen Estimator with Inequality Constraints for Expensive Hyperparameter Optimization

# The requirement is only Optuna.
$ pip install optuna

# You can also optionally install as follows:
$ pip install -r https://hub.optuna.org/samplers/ctpe/requirements.txt

from __future__ import annotations

import numpy as np
import optuna
import optunahub


def objective(trial: optuna.Trial) -> float:
    x = trial.suggest_float("x", 0.0, 2 * np.pi)
    y = trial.suggest_float("y", 0.0, 2 * np.pi)
    return float(np.sin(x) + y)


def constraints(trial: optuna.trial.FrozenTrial) -> tuple[float]:
    x = trial.params["x"]
    y = trial.params["y"]
    c = float(np.sin(x) * np.sin(y) + 0.95)
    trial.set_user_attr("c", c)
    # For multiple constraints, we can simply include every constraint in the tuple below.
    # e.g., if c1 <= t1 and c2 <= t2 are the constraints, return (c1 - t1, c2 - t2).
    return (c, )


sampler = optunahub.load_module(package="samplers/ctpe").cTPESampler(constraints_func=constraints)
study = optuna.create_study(sampler=sampler)
study.optimize(objective, n_trials=30)
print(study.best_trials)

@inproceedings{watanabe_ctpe_ijcai_2023,
    title={{c-TPE}: Tree-structured {P}arzen Estimator with Inequality Constraints for Expensive Hyperparameter Optimization},
    author={Watanabe, Shuhei and Hutter, Frank},
    booktitle={International Joint Conference on Artificial Intelligence},
    year={2023}
}

@inproceedings{watanabe_ctpe_workshop_2022,
    title={{c-TPE}: Generalizing Tree-structured {P}arzen Estimator with Inequality Constraints for Continuous and Categorical Hyperparameter Optimization},
    author={Watanabe, Shuhei and Hutter, Frank},
    journal = {Gaussian Processes, Spatiotemporal Modeling, and Decision-making Systems Workshop at Advances in Neural Information Processing Systems},
    year={2022}
}

pip install pytest

pytest package/samplers/ctpe/tests/