A exploratory side project to throw machine learning at a long-time favorite game of mine, Mahjong.
Just as a warning, the purpose isn't to produce an incredibly accurate ML model - that said, a swap to a more appropriate policy gradient method is probably necessary at some point, seeing as it plays questionably (at best).
Also as an warning, the current UI - I'll just say it's not the most user friendly.
Mahjong images courtesy of https://github.com/FluffyStuff/riichi-mahjong-tiles.git
RL algorithm design based pretty much entirely on "Reinforcement Learning: An Introduction (2nd Edition)" (R. Sutton and A. Barto)