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. COMBO 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 our document.
- Python 2.7.x
- numpy >=1.10
- scipy >= 0.16
- Cython >= 0.22.1
- mpi4py >= 2.0 (optional)
1. Download or clone the github repository, e.g.
> git clone https://github.com/tsudalab/combo.git
2. Run setup.py install
> cd combo
> python setup.py install
1. Delete all installed files, e.g.
> python setup.py install --record file.txt
> cat file.txt | xargs rm -rvf
After installation, you can launch the test suite from 'examples/grain_bound/tutorial.ipynb'.
This package is distributed under the MIT License.