The aim is to rise to the challenge offered by Deepmind [1].
One promising lead is offered by Facebook [2].
See the instructions in environment/README.md
.
If you run into the error:
ModuleNotFoundError: No module named 'skbuild'
you need to install scikit-buildCMake error
, try runningenvironment/clean_all.sh
then try again
Our first main goal is to rise to the first challenge, which is when the game happens with the same agent playing with himself.
Then, we'll move on to the next challenge which is to play with different agents.
+ environment/ contains Deepmind's hanabi environment
+ players/ contains the players we made
With equal contributions:
- Guillaume Grosse
- Matéo Mahaut
- Théo Matricon
- Quentin Lanneau
-
[1] The Hanabi challenge: A new frontier for AI research
Bard, Nolan and Foerster, Jakob N. and Chandar, Sarath and Burch, Neil and Lanctot, Marc and Song, H. Francis and Parisotto, Emilio and Dumoulin, Vincent and Moitra, Subhodeep and Hughes, Edward and et al.
from Artificial Intelligence vol. 20, March 2020, http://dx.doi.org/10.1016/j.artint.2019.103216
DOI: 10.1016/j.artint.2019.103216 -
[2] Improving Policies via Search in Cooperative Partially Observable Games
Adam Lerer and Hengyuan Hu and Jakob Foerster and Noam Brown
2019, https://arxiv.org/abs/1912.02318