DiversePolicies

Code accompanying the paper "Diverse Policies Converge in Reward-free Markov Decision Processes" (PRICAI 2023) [paper]

This repository provides visualization code for both the Policy Evolution Trajectory and the Policy Evolution Process. They can be run through the commands python draw_trajectory.py and python draw_evolution.py, respectively.

Cite

Please cite our paper if you use our codes or our weights in your own work:

@article{lin2023diversepolicies,
  title={Diverse Policies Converge in Reward-free Markov Decision Processes},
  author={Lin, Fanqi and Huang, Shiyu and Tu, Wei-Wei},
  journal={arXiv preprint arXiv:2308.11924},
  year={2023}
}