alpha-zero-general

Why we included this project

Few repositories make AlphaGo Zero's ideas as easy to follow as this one. The code is deliberately simplified and well commented, so you can trace how the neural network, Monte Carlo Tree Search, and self-play training loop fit together without wading through a production-scale codebase. Game logic stays separate from the network implementation, so you can add your own two-player turn-based game and switch between PyTorch and Keras with only a few lines changed. A pretrained Othello model and a written tutorial let you check the results hands-on, which is useful when the goal is building intuition before tackling bigger systems.