Go Go! - Evaluating Different Variants of Monte Carlo Tree Search for Playing Go
Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesis
Monte Carlo Tree Search (MCTS) is a Go algorithm that is used in many recent strong go playing agents. In this report we test and compare different algorithms related to Monte Carlo simulation, seeing how well they do against each other under different time constraints on consumer hardware. This is done with our own implementation of Go rules and algorithms written in Java. The All Moves As First (AMAF) algorithmhad the best performance in the performed tests.
Place, publisher, year, edition, pages
IdentifiersURN: urn:nbn:se:kth:diva-157520OAI: oai:DiVA.org:kth-157520DiVA: diva2:770360