Application of Machine Learning in Games with incomplete information

Abstract:
Machine Learning is applied to every aspect of people’s life. Especially in the field of games, machines gradually show their own set of mechanisms and occupy a certain position in this field. In 2016, Α Go and Lee Sedol, the world champion of Go and a professional nine-dan player, had a man-machine battle with Go and won with a total score of 4-1. However generally, games such as Go and Chess are classified as games with complete information, while other games, such as poker, are games with incomplete information. In the game of incomplete information, traditional algorithms are no longer applicable due to the lack of information and the increase in uncertainty, and an independent algorithm system needs to be studied. This paper first introduces the application of machine learning in known-complete information games, then explains the problems of machine learning in multi-players known incomplete information games, summarizes the development of machine learning in the field of incomplete information games at this stage, and finally prospects the future of machine learning in the field of games. In general, machine learning still has problems in the game of incomplete information, and future generations need to continue to study.