The basis of the method proposed in this article is the idea that information is one of the most important factors in strategic decisions, including decisions in computer chess and other strategy games. The model proposed in this article and the algorithm described are based on the idea of a information theoretic basis of decision in strategy games . The model generalizes and provides a mathematical justification for one of the most popular search algorithms used in leading computer chess programs, the fractional ply scheme. However, despite its success in leading computer chess applications, until now few has been published about this method. The article creates a fundamental basis for this method in the axioms of information theory, then derives the principles used in programming the search and describes mathematically the form of the coefficients. One of the most important parameters of the fractional ply search is derived from fundamental principles. Until now this coefficient has been usually handcrafted or determined from intuitive elements or data mining. There is a deep, information theoretical justification for such a parameter. In one way the method proposed is a generalization of previous methods. More important, it shows why the fractional depth ply scheme is so powerful. It is because the algorithm navigates along the lines where the highest information gain is possible. A working and original implementation has been written and tested for this algorithm and is provided in the appendix. The article is essentially self-contained and gives proper background knowledge and references. The assumptions are intuitive and in the direction expected and described intuitively by great champions of chess.
Charles University(查理大学);University of Alberta(阿尔伯塔大学)
从这个领域的黎明开始,带有价值函数的搜索就是计算机游戏研究的一个基本概念。图灵在1950年的国际象棋算法能够提前两步思考,香农在$1950$的国际象棋工作包括一个关于搜索中使用的评估函数的广泛部分。Samuel在1959年开发的跳棋程序已经将搜索和价值函数结合在一起,这两个函数都是通过自我游戏和自举学习的。TD-Gammon改进了这些想法,并使用神经网络来学习那些复杂的价值函数--只是在搜索中再次使用。决策时间搜索和价值函数的结合已经出现在计算机在长期挑战性游戏中击败人类对手的显着里程碑中- DeepBlue for Chess和AlphaGo for Go。直到最近,这个强大的搜索框架与(学习)价值函数的帮助下,已被限制在完美的信息游戏。由于许多有趣的问题并不能为智能体提供环境的完美信息,这是一个不幸的限制。本文介绍了不完全信息博弈下的声音搜索.