Tic-Tac-Toe AI using Deep Q-Learning

On December 5, 2017, DeepMind introduced AlphaZero to the world, a chess-playing program that defeated Stockfish simply by playing against itself for 4 hours. Impressed by AlphaZero, I wrote a simple AI to play Tic-Tac-Toe based on Deep Q-Learning, and today I want to share it with you. In this article, I will cover Markov Decision Process, Q-Learning, Deep Q-Learning, and the application of Deep Q-learning to Tic-Tac-Toe.