A Research on Some Upper Limits in Artificial Intelligence of the Game of Go
|School||Central South University|
|Course||Probability Theory and Mathematical Statistics|
|Keywords||the game of Go live groups of stones Monte-Carlo UCT complexity|
The artificial intelligence of the game of Go develops quickly. The core of the computer Go is Monte-Carlo and UCT algorithm when they made some success. It is the main idea of this paper to find some upper limits in artificial intelligence research of the game of Go. It introduces the base of complexity’s degradation. It tells us how to find some upper limits in the game of Go, degrade the complexity of the game of Go effectively, and judge the type of local shape of Go. We find a fatal defect of UCT and Monte-Carlo through an instance. It is necessary to find the inherent rules of the game of Go. I try my best to find all kinds of upper limits. We have proved the upper limit of the live groups of black stones is20and the number of the live groups of stones is no more than33when we don’t consider seki. The upper limit of the nested layers is19. A strong hypothesis is presented that the number of live groups will not exceed71.