Evolving the MCTS Upper Confidence Bounds for Trees Using a Semantic-inspired Evolutionary Algorithm in the Game of Carcassonne

Monte Carlo Tree Search (MCTS) is a sampling best-first method to search for optimal decisions. The success of MCTS depends heavily on how the tree built and selection process plays fundamental role in this. One particular mechanism that has proved be reliable based Upper Confidence Bounds Trees (UCT). UCT attempts balance exploration exploitation by considering values stored statistical MCTS. However, some tuning necessary this work well. In work, we use Evolutionary Algorithms (EAs) evolve mathematical expressions with goal substitute formula evolved More specifically, means our proposed Semantic-inspired Algorithm approach (SIEA-MCTS). This inspired semantics Genetic Programming (GP), where fitness cases seen as requirement adopted GP. Fitness are normally used determine individuals can compute semantic similarity (or dissimilarity) individuals. not available We extend notion using multiple reward from allow us both an individual its semantics. By doing so, show SIEA-MCTS able successfully yield better or competitive results compared without need these expressions. compare performance against algorithms, Rapid Action Value Estimation three variants *-minimax family random controller two more EA approaches. consistently outperforms most intelligent controllers challenging game Carcassonne, whose state-space complexity is, approx., 10 $^{40}$ .