In decision making our current knowledge plays a big part, especially when thinking of possibilities that could happen in the future. For humans, long-term planning is hard, since it is, among other reasons, hard to remember. In decision-making problems, which include also games, we want to perform such actions that most probably lead to success. One of such methods is Monte Carlo tree search. Because it is too hard for humans to use it on their own, we try to implement it using computers, since it is a method that works well on games with large branching factors.
In this thesis we implement the method and show how to use it in a game called Nine men's morris, which is moderately complex and so it is hard to predict it into the future.
|
