ozzi7 / TicTacToeRL

A simple implementation of Q-learning for Tic Tac Toe

TicTacToeRL

A simple implementation of Q-learning for Tic Tac Toe.

Q-learning parameters:
Alpha = 0.9, Learning rate
Gamma = 1.0, Discount rate
Epsilon = 0.8, Probability for random moves

Converges in approximately 50 000 training games.