hantaoyin/mcts-go

An AlphaGo imitation.

mcts-go

An AlphaGo imitation.

Reference:

1. The AlphaGo Zero paper: Mastering the game of Go without human knowledge, Nature volume 550, pages 354–359 (19 October 2017).

2. leela-zero (https://github.com/leela-zero/leela-zero, an AI Go player using the same reinforcement learning approach as AlphaGo): Some ideas (but not code) about how to implement the Go game engine are borrowed from it.

This implementation is intended for educational purposes. It can only run on a single machine and will probably be too slow for practical purposes on any Go board size larger than 9x9. The author trained a model on a 9x9 Go board using this implementation on a machine with Nvidia GeForce GTX 1080 Ti in about a month. If you are interested in a more practical implementation that can play the full size 19x19 Go game, check out leela-zero as mentioned above.