The Go Death And Life Knowledge Discovery System Based On Logic Reasoning And Machine Learning

This paper presents an explanation-based learning system, with the instruction of life and death knowledge in Go, and combining with computer logic reasoning, we discover and extract life and death knowledge, which can be used by computer. In usual, that knowledge is summarized by hand, so there is the quantity problem, which can be resolved well by the method this paper represents.Computer Go is a difficulty in AI domain. Owing to Go the particularity of Go, some search algorithms successful used in chess cannot use in Go successfully. Effective and express Go directly mathematic model is not able to establish also, so at present level of the computer Go is much low, even can not reach low dan professional player's level。In the process of searching other approach, the human player's thinking method used in studding playing Go or playing Go has catch AI domain's attention. One three or five years old child, through study can learn how to play Go in short time, we know that we learn something mainly by imitation and memory; can we learn something about this? Now there is plenty of Go tutorial and Go study material, how to make computer to use those material? This is meaningful to the developing of AI and the discussion of human intelligence indeed.This paper presents an explanation-based learning system, with the instruction of life and death knowledge in Go, and combining with computer logic reasoning, we discover and extract life and death knowledge, which can be used by computer. In usual, that knowledge is summarized by hand, so there is the quantity problem, which can be resolved well by the method this paper represents.The main research work in this paper as follow:①Divide Go Knowledge into distinct catalogs, and treat them respectively. For that regularization knowledge we use first order predicate logic; for those knowledge relied on shape we use shape patterns. Besides, there is operation knowledge that generate move directly.② At the base of first order predicate logic this paper construct a strategy logic reasoning machine model, and discuss the correctness of this model.③Give out a explanation based machine learning method and how to apply it into extracting computer Go life-death knowledge.④with those base theory, We give out the whole system model ,data structure, algorithms and design details, at last we will discuss something about test.⑤Give out the detail methods used in Computer Go life-death knowledge representation, and discuss how to organize and use those knowledge also.