abstract
- Bayesian optimization has emerged as a powerful tool for efficiently exploring complex, data-driven design spaces. We introduce a novel Bayesian optimization library incorporating superlevel set filtration for space partitioning, enabling the algorithm to focus on regions of interest. This approach aims to enhance the balance between exploration and exploitation in Bayesian optimization, enabling an effective strategy for evaluating multiple points simultaneously. Additionally, the library features automatic kernel selection to improve the accuracy and robustness of the surrogate model by optimally fitting the data without requiring manual intervention. We also propose methods to manage and model constraint functions when the optimization problem requires it. We tested the library using three challenging case studies, where we explored the implementation details and demonstrated its potential through reference examples. Our findings indicate that our method outperforms standard Bayesian optimization, offering a robust and user-friendly solution for addressing complex and computationally expensive optimization problems. © 2025 American Chemical Society.