mctslib: Reproducible Real-time Statistical Decision-making

Monte Carlo tree search (MCTS) is a search algorithm which has used to achieve state-of-the-art performance in many domains. The algorithm’s generality, effectiveness in complicated and large environments, and relative simplicity has made it one of the foremost search methods. In addition, MCTS and related algorithms can often be easily modified to suit a specific domain. Despite their usefulness, public implementations of these algorithms are rare, unlike other subareas of AI like deep learning or reinforcement learning. Moreover, the implementations that are public are often domain-specific, inefficient, difficult to use, and aren’t fully featured. This thesis aims to provide robust and easy to use implementations of tree search algorithms, as well as an analysis of how the characteristics of problem domains impact the effectiveness of different search algorithms.