This project contains code for the paper titled "Reinforcement Learning When All Actions are Not Always Available" that appeared at the Thirty-fourth Conference on Artificial Intelligence (AAAI 2020).

Pdf: https://arxiv.org/abs/1906.01772

This code is written in Python 3.6

Environment This folder contains the maze environment used in the paper.

Src This folder contains the code for the proposed and baseline algorithms.

To run the code:

Src/SAS_parser.py First set the desired hyper-parameters in this file.

Src/run_SAS.py After that execute this file.

Note: You might need to appropriately set the root of your project for the import commands to work.

@inproceedings{chandak2020reinforcement,
  title={Reinforcement Learning When All Actions Are Not Always Available.},
  author={Chandak, Yash and Theocharous, Georgios and Metevier, Blossom and Thomas, Philip S},
  booktitle={AAAI},
  pages={3381--3388},
  year={2020}
}