belakaria/hybo

Bayesian Optimization over Hybrid Spaces

This repository contains the source code for the paper "Bayesian Optimization over Hybrid Spaces" presented at Thirty-eighth ICML'21 conference.

• By default, data is stored in ../EXPERIMENTS. Directory can be changed in config.py

• The command-line arguments are described below:

  • n_eval : The number of evaluations
  • objective : ['coco'] (example on how to create new objective can be seen in experiments/test_functions/mixed_integer.py)
  • problem_id : applicable only for 'coco' and 'nn_ml_datasets' domain
  • path : A path to the directory of the experiment to be continued. (Only required when you want to resume an experiment)

• Example run python main.py --objective coco --problem_id bbob-mixint_f001_i01_d10 --n_eval 180

• There are 7 benchmarks used in the paper. For using synthetic benchmark, please see instructions in coco suite. For robot pushing benchmark, please see the original description in Ensemble-Bayesian-Optimization.

The discrete part of the code is built upon the source code provided by the COMBO authors. We thank them for their code and have added appropriate licenses.