Regional Expected Improvement for Efficient Trust Region Selection in High-Dimensional Bayesian Optimization

This repository provides the source code proposed in the following article. Please cite this article if you use the code.

Nobuo Namura and Sho Takemori, "Regional Expected Improvement for Efficient Trust Region Selection in High-Dimensional Bayesian Optimization," In Proceedings of the 39th AAAI Conference on Artificial Intelligence (2025). [arXiv]

Requirements

To install requirements:

pip install -r requirements.txt

Usage

1. Install the libraries listed in requirements.txt (see above).
2. Edit the main.py for your setting.
3. Run python main.py.

• To use the HPA problems, you need to download or "git clone" this repository, and place it into src/benchmark/hpa
  • Directory tree must be as follows

    regional-expected-improvement
    └── src/
        └── benchmark/
            ├── ebo
            ├── hpa/
            │   ├── hpa/
            │   │   ├── airfoil_info
            │   │   ├── __init__.py
            │   │   ├── adapter.py
            │   │   ├── designer.py
            │   │   └── problem.py
            │   ├── igd_reference_points
            │   ├── img
            │   └── utopia_and_nadir_points
            └── mopta08
    

• To use the MOPTA08 problem, you need to download binaries from here, and place them into src/benchmark/mopta08

License

• This project is under the MIT License. See LICENSE for details.