ablancha / gpsearch

Bayesian optimization and active learning with likelihood-weighted acquisition functions

gpsearch

Source code for Bayesian optimization and active learning with likelihood-weighted acquisition functions.

Installation

Clone the repo, then create a fresh conda environment from the requirements file and install using pip.

git clone https://github.com/ablancha/gpsearch.git
cd gpsearch
conda create --name myenv --file requirements.txt -c conda-forge -c dmentipl
conda activate myenv
pip install .

Notes

The acquisition functions available in gpsearch are compatible with 1.9.9 of GPy. Beware of this issue if you decide to use a different version.

Benchmarks

The following benchmarks are included:

• stochastic oscillator (used here)
• extreme-event precursor (used here and here)
• borehole function (used here)
• synthetic test functions (used here)
• brachistochrone problem (unpublished)

References

• Bayesian Optimization with Output-Weighted Optimal Sampling
• Informative Path Planning for Extreme Anomaly Detection in Environment Exploration and Monitoring
• Output-Weighted Optimal Sampling for Bayesian Experimental Design and Uncertainty Quantification