Skip to content
/ bae Public

Code for the paper: Bootstrap Advantage Estimation for Policy Optimization in Reinforcement Learning. https://arxiv.org/abs/2210.07312

Notifications You must be signed in to change notification settings

masud99r/bae

Repository files navigation

Bootstrap Advantage Estimation (BAE)

Paper Title: Bootstrap Advantage Estimation for Policy Optimization in Reinforcement Learning. (ICMLA 2022). arXiv

Installation

This codebase is based on the CleanRL library. Please follow the installation instructions on the official library site. The code should run on version CleanRL v1.0.0b1. It might work on the latest version; however, some modifications might be needed.

The Procgen environments are based on image-based observation, and the policy and value networks use CNN-based models. Thus it is recommended to run those experiments on a GPU-enabled machine.

The experiments on DeepMind Control and PyBullet environments should run on the CPU.

For DeepMind Control environments, we use dmc2gym to convert the environments to OpenAI Gym format to be compatible with CleanRL. Please follow the instruction in the original repository for installation. A copy of the repository is provided here as well.

Contact

Please contact the author at [email protected] if you have any queries.

Citation

If you use this code or data, please consider citing this paper:

@inproceedings{rahman2022bae,
  title={Bootstrap Advantage Estimation for Policy Optimization in Reinforcement Learning},
  author={Rahman, Md Masudur and Xue, Yexiang},
  booktitle={IEEE International Conference on Machine Learning and Applications (ICMLA 2022)},
  year={2022}
}

About

Code for the paper: Bootstrap Advantage Estimation for Policy Optimization in Reinforcement Learning. https://arxiv.org/abs/2210.07312

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages