/BOPAH

The official implementation of "Batch Reinforcement Learning with Hyperparameter Gradients", ICML 2020

Primary LanguagePythonMIT LicenseMIT

Batch Reinforcement Learning with Hyperparameter Gradients

This repository is the official implementation of Batch Reinforcement Learning with Hyperparameter Gradients.

Requirements

To install requirements:

conda env create -f environment.yml
conda activate batchrl

To download the batch trajectories used in the paper, please run the following:

python download_dataset.py

Finite MDP experiments

To run the finite MDP experiments in the paper, run this command:

python finite_run.py

Continuous control experiments

To run the Mujoco continuous control experiments in the paper, run this command:

python cont_run.py

References

If this repository helps you in your academic research, you are encouraged to cite our paper. Here is an example bibtex:

@inproceedings{lee2020batch,
	title={Batch Reinforcement Learning with Hyperparameter Gradients},
	author={Byung-Jun Lee* and Jongmin Lee* and Peter Vrancx and Dongho Kim and Kee-Eung Kim},
	booktitle={Proceedings of the 37th International Conference on Machine Learning},
	year={2020}
}