scikit-optimize

A Python library for sequential model-based optimization designed to minimize expensive, noisy black-box functions.

Scikit-optimize (skopt) streamlines hyperparameter tuning through Bayesian optimization techniques like Gaussian processes and decision trees. It serves as a direct upgrade to traditional grid or random search methods by using past evaluation results to inform future trials. The library integrates seamlessly with scikit-learn via the BayesSearchCV class: users replace standard search CVs with a few lines of code to find optimal parameters in fewer iterations. It supports complex search spaces (including categorical and integer dimensions) and provides built-in visualization tools to analyze convergence and objective landscapes.

https://scikit-optimize.github.io/stable/