Reinforcement learning using Monte Carlo Tree Search

A collection of games were agents use the Monte Carlo Tree Search algorithm to play against each other. The agents don't require any human supervision and will learn from self-play given the game rules.

Playing a game of tic tac toe

Two agents play a game by simulating their next move with a UCT (Upper Confidence Bound 1 applied to trees) score. The agents are capable of solving the game and the only winning move is apparently not to play.

Supported games

• Tic tac toe

Future development

• Add more games.
• Implement ANNs.