addyg/QLearning-TicTacToe

Using QLearning to play a game of Tic-Tac-Toe against 3 types of opponents: Random, Smart, and Perfect Players

QLearning-TicTacToe

Using QLearning to play a game of Tic-Tac-Toe against 3 types of opponents: Random, Smart, and Perfect Players

6 Files:

• Board.py: code to make tic-tac-toe board, 3 by 3 grid
• RandomPlayer: moves randomly
• SmartPlayer: better than RandomPlayer, but cannot beat PerfectPlayer
• PerfectPlayer: never loses
• QLearner.py: Q-Learning Player Code
• TicTacToe.py: driver code, where all players play tic-tac-toe games and where our QLearner is trained and tested

Q-Learning:

Q ( s , a ) ← (1 - ⍺) Q ( s , a ) + ⍺ ( R ( s ) + 𝛾 max a ’ Q ( s’ , a ’))

Parameters:

Reward for WIN,DRAW,LOSE, learning rate, discount factor and initial conditions.