A simplified implementation of DeepMind's AlphaZero (with Monte Carlo Tree Search, etc), which can be used to train agents to play simple board games, such as Connect4, Tic-Tac-Toe, and Five in a Row (Gomoku), etc.

Implementation

Connect 4 Board and Rules

Create a board

Self Play

Use Monte Carlo Tree Search with self play

References:

• AlphaZero for Connect4 in PyTorch, Wee Tee Soh, April 2019 https://towardsdatascience.com/from-scratch-implementation-of-alphazero-for-connect4-f73d4554002a
• AZFour for Connect4 (6 parts), by Anthony Young, Sept 2019, https://medium.com/@sleepsonthefloor/azfour-a-connect-four-webapp-powered-by-the-alphazero-algorithm-d0c82d6f3ae9
• AlphaZero Connect4 with Keras, David Foster, Jan 2019 https://medium.com/applied-data-science/how-to-build-your-own-alphazero-ai-using-python-and-keras-7f664945c188