optimization tools for PHYsics based on Bayesian Optimization ( PHYSBO )

Bayesian optimization has been proven as an effective tool in accelerating scientific discovery. A standard implementation (e.g., scikit-learn), however, can accommodate only small training data. PHYSBO is highly scalable due to an efficient protocol that employs Thompson sampling, random feature maps, one-rank Cholesky update and automatic hyperparameter tuning. Technical features are described in COMBO's document and PHYSBO's report (open access). PHYSBO was developed based on COMBO for academic use.

Document

• Stable (master branch)
  • English
  • 日本語
• Latest (develop branch)
  • English
  • 日本語

Required Packages

• Python >= 3.6
  • No longer tested with Python 3.6
• NumPy < 2.0.0
• SciPy

Install

• From PyPI (recommended)

python3 -m pip install physbo

• From source (for developers)

  1. Update pip (>= 19.0)

     python3 -m pip install -U pip

  2. Download or clone the github repository

     git clone https://github.com/issp-center-dev/PHYSBO
     

  3. Install via pip

     # ./PHYSBO is the root directory of PHYSBO
     # pip install options such as --user are avaiable
     
     python3 -m pip install ./PHYSBO

  4. Note: Do not import physbo at the root directory of the repository because import physbo does not try to import the installed PHYSBO but one in the repository, which includes Cython codes not compiled.

Uninstall

python3 -m pip uninstall physbo

Usage

'examples/simple.py' is a simple example.

Data repository

A tutorial and a dataset of a paper about PHYSBO can be found in PHYSBO Gallery.

License

PHYSBO was developed based on COMBO for academic use. PHYSBO v2 is distributed under Mozilla Public License version 2.0 (MPL v2). We hope that you cite the following reference when you publish the results using PHYSBO:

“Bayesian optimization package: PHYSBO”, Yuichi Motoyama, Ryo Tamura, Kazuyoshi Yoshimi, Kei Terayama, Tsuyoshi Ueno, Koji Tsuda, Computer Physics Communications Volume 278, September 2022, 108405.

Bibtex

@misc{@article{MOTOYAMA2022108405,
title = {Bayesian optimization package: PHYSBO},
journal = {Computer Physics Communications},
volume = {278},
pages = {108405},
year = {2022},
issn = {0010-4655},
doi = {https://doi.org/10.1016/j.cpc.2022.108405},
author = {Yuichi Motoyama and Ryo Tamura and Kazuyoshi Yoshimi and Kei Terayama and Tsuyoshi Ueno and Koji Tsuda},
keywords = {Bayesian optimization, Multi-objective optimization, Materials screening, Effective model estimation}
}

Copyright

© 2020- The University of Tokyo. All rights reserved. This software was developed with the support of "Project for advancement of software usability in materials science" of The Institute for Solid State Physics, The University of Tokyo.