vanderschaarlab/optcommit

When to make and break commitments?

When to Make and Break Commitments?

Code author: Alihan Hüyük (ah2075@cam.ac.uk)

This repository is for reproducing the main experimental results in the ICLR'23 paper "When to make and break commitments?" The method we propose, Bayes-OCP, is implemented as the function policy_bayesocp in src/algs.py.

Usage

First, clone the repository, then install the required python packages by running:

python -m pip install -r requirements.txt

Then, the main results presented in Table 3 can be reproduced by running:

python src/main.py
python src/eval.py > res.txt

Notes

• Note there are empty cells in the results table you will see when running python src/eval.py for Oracle RCT. These results are determined by the problem setting directly (and they do not need to be computed numerically through simulations).
• Note that for some folds, Greedy Bayes OCP never succeeds (hence it is not possible to compute time-to-success and you may see a nan value in the python src/eval.py results table). The paper reports the average (and variance) for folds where Greedy Bayes OCP succeeded at least once.

Contributing

If you would like to contribute to the code, please install and enable pre-commit before making any commits and submitting a PR:

python -m pip install pre-commit
pre-commit install

The checks pre-commit will run can be found in .pre-commit-config.yaml. You will need to fix any problems identified by these checks before you can commit your changes.

Citing

If you use this software, please cite as follows:

@inproceedings{huyuk2023when,
  author={Alihan H\"uy\"uk and Zhaozhi Qian and Mihaela van der Schaar},
  title={When to make and break commitments?},
  booktitle={Proceedings of the 11th International Conference on Learning Representations},
  year={2023}
}