Dissertation
Title: Guiding Monte Carlo Tree Search simulations through Bayesian Opponent Modeling in The Octagon Theory
Author: Hugo Fernandes
Adviser: Eugénio Oliveira (PhD)
Co-Adviser: Pedro Nogueira (MSc)
Institute: LIACC - Artificial Intelligence and Computer Science Laboratory
Summary : This thesis builds on the current state-of-the-art in Monte Carlo Tree Search (MCTS) methods, by investigating the integration of Opponent Modeling with MCTS. The goal of this integration is to guide the simulations of the MCTS algorithm according to knowledge about the opponent, obtained in real-time through Bayesian Opponent Modeling, with the intention of reducing the number of irrelevant computations that are performed in purely stochastic, domain-independent methods. For this research, the two player deterministic board game The Octagon Theory was used.
Deliverables: Please click here and insert your credentials.