Practical Bayesian Algorithm Execution via Posterior Sampling

Practical Bayesian Algorithm Execution via Posterior Sampling
Chu Xin Cheng, Raul Astudillo, Thomas Desautels, Yisong Yue
Advances in Neural Information Processing Systems, 2024

Environment setup

This project uses conda to manage the environment. Clone the repository and navigate to the root directory.

$ git clone https://github.com/RaulAstudillo06/PSBAX.git
$ cd PSBAX

To set the environment, run the following command:

$ conda env create -f env.yml

Activate the environment:

$ conda activate PSBAX

Demos

In the demos directory, there are three demo files that can be run to showcase the capabilities of PSBAX.

• Demo 1: Local Bayesian Optimization

  $ python local_bo.py
  

• Demo 2: Level Set Estimation

  $ python level-set.py
  

• Demo 3: Top-k Estimation

  $ python topk.py
  

The experiment results are saved in the demos/results directory. A graph of the results is automatically generated and saved in the demos/plots directory.

Main experiments

The experiments directory contains the runner files used to run the experiments conducted in the main paper.

DiscoBAX

The DiscoBAX experiments were performed using data from the paper "DiscoBAX - Discovery of optimal intervention sets in genomic experiment design" by Lyle et al. The data included in this repository is preprocessed and saved in the discobax/data directory. To access the full dataset, please refer to the original DiscoBAX repository here.

To run the experiments for DiscoBAX, navigate to the experiments/discoBAX directory and run the following command:

$ bash run_discobax.sh discobax_runner.py

Local Bayesian Optimization

To run the experiments for Local Bayesian Optimization, navigate to the experiments/single-objective directory and run the following commands:

• Ackley 10D synthetic function:

  $ bash run_local-bo.sh ackley_runner.py
  

• Hartmann 6D synthetic function:

  $ bash run_local-bo.sh hartmann_runner.py
  

Level Set Estimation

To run the experiments for Level Set Estimation, navigate to the experiments/level-set directory and run the following command:

$ bash run_level-set.sh

Top-k Estimation

To run the experiments for Top-k Estimation, navigate to the experiments/topk directory and run the following commands:

• Rosenbrock synthetic function:

  $ bash run_topk.sh topk_runner.py
  

• GB1 protein fitness:

  $ bash run_topk.sh gb1_runner.py
  

Citation

@inproceedings{neiswanger2021bayesian,
  title         = {Practical Bayesian Algorithm Execution via Posterior Sampling},
  author        = {Cheng, Chu Xin and Astudillo, Raul and Desautels, Thomas and Yue, Yisong},
  booktitle={Advances in Neural Information Processing Systems},
  volume = {38},
  publisher = {Curran Associates, Inc.},
  year={2024}
}