Maximum Entropy Reinforcement Learning via Energy-Based Normalizing Flow

<br>

This repository contains the code implementation of the experiments presented in the paper Maximum Entropy Reinforcement Learning via Energy-Based Normalizing Flow.

---

Directory Structure

• Use the code in meow/toy to reproduce the experimental results presented in Section 4.1 of our paper.
• Use the code in meow/cleanrl to reproduce the experimental results presented in Section 4.2 of our paper.
• Use the code in meow/skrl to reproduce the experimental results presented in Section 4.3 of our paper.
• Use the code in meow/plot to reproduce the figures presented in our paper.

---

License

To maintain reproducibility, we freezed the released versions of following repositories and list their licenses as follows:

• Toni-SM/skrl (at commit 631613a) is licensed under the MIT License.
• vwxyzjn/cleanrl (at commit 8cbca61) is licensed under the MIT License.
• VincentStimper/normalizing-flows (at commit 848277e) is licensed under the MIT License.
• rail-berkeley/softlearning (at commit 13cf187) is licensed under the MIT License.

Further changes based on the repository above are licensed under the MIT License.

---

Cite this Repository

If you find this repository useful, please consider citing our paper:

@inproceedings{chao2024maximum,
    title={Maximum Entropy Reinforcement Learning via Energy-Based Normalizing Flow},
    author={Chao, Chen-Hao and Feng, Chien and Sun, Wei-Fang and Lee, Cheng-Kuang and See, Simon and Lee, Chun-Yi},
    booktitle={Proceedings of the International Conference on Neural Information Processing Systems (NeurIPS)},
    year={2024}
}

---

Contributors of the Code Implementation

<img src="toy/src/MEow_c.gif" alt="meow" width="10%" align="center"> <img src="toy/src/MEow_wf.gif" alt="meow" width="10%" align="center"> <img src="toy/src/MEow_ch.gif" alt="meow" width="10%" align="center">

Visit our GitHub pages by clicking the images above.