Monte Carlo Tree Search

Exploration and Exploitation: Machine Learning for Games

I recently saw the AlphaGo documentary film and it blew my mind. Essentially, a computer was able to beat a world champion in a best of five match of Go, a game with an overwhelmingly large search space—a 19 x 19 board. One of the principal techniques of its success was Monte Carlo Tree Search (MCTS). While it has received adequate exposition elsewhere,1 I thought that a video could be a nice addition to what’s already out there.