A Go program with no human provided knowledge. Using MCTS (but without Monte
Carlo playouts) and a deep residual convolutional neural network stack.

This is a fairly faithful reimplementation of the system described in the Alpha
Go Zero paper "Mastering the Game of Go without Human Knowledge". For all
intents and purposes, it is an open source AlphaGo Zero.