Trying to implement a simplified version of AlphaGo (by "Google DeepMind") – a computer program that uses the Monte-Carlo tree search algorithm combined with a (not so) deep Convolutional Neural Network in order to play Connect4.
- board.py : A collection of classes and functions for playing certain types of games.
- connect_four.py : A simulator for playing Connect4 with different players (mAlphaGo, standard MCTS, random, human).
- train_cnn.py : building and training the cnn (the 'mAlphaGo' player) using Tensorflow.
(disclaimer - The code is quite messy, largely uncommented, and intended mainly for reference. I've written it when I was just starting with python and ML. Further work will hopefully be better)