MCMC-BO: Improving sample efficiency of high dimensional Bayesian optimization with MCMC on approximated posterior ratio

Overview

MCMC-BO is a Bayesian optimization method that leverages Markov Chain Monte Carlo to efficiently sample from an approximated posterior by guiding candidate points toward more promising regions of the search space.

This repository provides an implementation of the MCMC-BO algorithm proposed in the following paper:

Improving Sample Efficiency of High-Dimensional Bayesian Optimization with MCMC
Zeji Yi, Yunyue Wei, Chu Xin Cheng, Kaibo He, Yanan Sui
Learning for Dynamics & Control Conference (L4DC), 2024
PDF

Installation

There is requirements.txt in this repository, which are mandatory to run MCMC-BO on synthetic functions.

If you want to run experiment on Mujoco locomotion tasks, please refer to https://github.com/openai/mujoco-py to configure corresponding environment.

How to run the algorithm

You can test MCMC-BO by running the following command:

python exp.py --func Ackley --dim 200  --tr_num 1 --eval_num 6000 --init_num 200 --batch_size 100 --noise_var 0 --repeat_num 30 --use_mcmc 1 --gpu_idx 0

BibTeX

@inproceedings{yi2024improving,
  title={Improving sample efficiency of high dimensional Bayesian optimization with MCMC},
  author={Yi, Zeji and Wei, Yunyue and Cheng, Chu Xin and He, Kaibo and Sui, Yanan},
  booktitle={6th Annual Learning for Dynamics \& Control Conference},
  pages={813--824},
  year={2024},
  organization={PMLR}
}