Skip to content

Connect 4 trained using MCTS + Deep Learning (AlphaGo Zero style)

License

Notifications You must be signed in to change notification settings

tonberry22/connect4-AlphaZero

Repository files navigation

connect4-AlphaZero

Connect 4 trained under a Reinforcement Learning paradigm, using MCTS + Deep Learning (AlphaGo Zero style), learning via pure self-play without feature engineering or expert input

Based on David Foster's code and model (https://github.com/AppliedDataSciencePartners/DeepReinforcementLearning)

Trained and tested under Windows 7 and Python 3.6

Feel free to clone and play a few games against the trained models (in the folder run0001, trained over 4 days using 1 Nvidia 1050 GPU) using the run.ipynb!

About

Connect 4 trained using MCTS + Deep Learning (AlphaGo Zero style)

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published