Towards Generalizing the Success of Monte-Carlo Tree Search beyond the Game of Go
Gusmão António, Tapani Raiko · Frontiers in artificial intelligence and applications · 2012
Monte-Carlo Tree Search and specifically the variants of the UCT algorithm have been a break-through in AI of the board game Go. However, UCT has had limited applicability to other domains. We study the limitations of some of the existing variants of UCT in a small-scale Markov decision process (MDP), and propose new variants that can reduce those limitations. Our experiments show great improvements in performance against traditional UCT and comparable performance to estabilished reinforcement learning algorithm, thus opening possibilities for applying UCT in other problem domains.