Parallel Monte Carlo Search for Imperfect Information Game Daihinmin

Computer games, such as chess, have recently become an important research topic in the field of computation intelligence. Monte Carlo massive searching is very successful in the game of GO as it has a huge intractable search space and no good partial evaluation heuristics. Imperfect information games also have intractable properties, such as large scale search space, less heuristics, and unknown state information. Monte Carlo search is therefore effective to make computer players for the imperfect information game. There is always a time limit for using computational intelligence in actual applications. Such constraints require parallel processing to be speed up the Monte Carlo simulations. We introduce a parallel Monte Carlo searching method for imperfect information card game Daihinmin, a familiar card game in Japan. Computer Daihinmin competitions have been held since 2006. We present two kinds of parallelization algorithms for Monte Carlo computer Daihinmin players.