Mastering Dots and Boxes with deep neural networks and tree search

This is the documentation for the code of the master’s thesis Mastering Dots and Boxes with deep neural networks and tree search. The code is split in two main parts: KSquares and AlphaDots

• KSquares provides an interface to play the game, hosts various AIs and provides facilities to generate data for training neural networks.
• AlphaDots is a collection of Python code and Jupyter notebooks that train, evaluate and host various neural networks.

There is a Docker Image to make it as easy as possible to reproduce the results and play Dots and Boxes with KSquares.