PipeBO: Pipelining Bayesian optimization

This is the code for Asynchronous Batch Bayesian Optimization with Pipelining Evaluations for Experimental Resource–constrained Conditions. This project is carried out in Funahashi Lab. at Keio University

Overview

PipeBO is a Bayesian optimizaiton method that performs parallelization through pipelining. Pipelining enables parallelization of experiments where batch Bayesian optimization cannot be applied due to equipement limitations. The implementation is based on GPyOpt [1].

Requirements

• Python 3.9
• coco-experiment
• Matplotlib
• NumPy <2.0.0
• SciPy <=1.12.0
• Pandas
• GPy

See requirements.txt for details

QuickStart

1. Download this repository by git clone.

   % git clone [email protected]:funalab/PipeBO.git

2. Install requirements.

   % cd PipeBO/
   % python -m venv venv
   % source ./venv/bin/activate
   % pip install --upgrade pip
   % pip install -r requiremets.txt
   % cd src/gpyopt
   % python setup.py develop

3. Run PipeBO with benchmark function.

   % cd PipeBO/src/
   % python main.py -bf f01_i01 -json ../confs/example.json  -rs 1 -iter 10

Reproduce the figure in the paper

Download data from here. The source code is in src/makefig/.

Acknowledgement

The development of this algorithm was funded by JST CREST (Grant Number JPMJCR21N1) to Akira Funahashi.

References

[1] The GPyOpt authors, GPyOpt: A Bayesian Optimizaiton framework in python (2016).