The aim of this repository is to demonstrate the use of machine learning/reinforcement learning techniques to simple games. I talk about these at https://projects.mcroci.com/
tic-tac-toe: Using temporal-difference learning to train a computer to play this classic game.
- Completed: Learning/training process for bot "X" vs bot "O"
- Completed: Testing bot "X" vs bot "O"
- Completed: Player versus bot as either "X" or "O" mode after testing
- Completed: Training and testing results analysis
ConnectX: Using temporal-difference learning to train a computer to play variants of connect 4.
- Completed: Connect3 on 4x4 board - player 1 can force win every game
- Completed: Connect4 on 6x7 board - 2M trials are not enough to train a skillful bot, a different RL/ML strategy must be used