Abstract:
This paper gives a few filling and deleting rules to imitate artificial intelligence to solve Sudoku under the candidate mode, and based on the proposed artificial intelligence rules, it constructs the difficulty scale and the algorithm of artificial intelligence solving. Validating the scale with a sample of 500 Dr.Sudoku puzzles rated externally into five gradations of difficulty. Goodman-Kruskal r coefficient is 0.82, indicating significant correlation the two scales. It gives an algorithm that can generate five different levels of Sudoku puzzles randomly.
Key words:
Sudoku,
filling and deleting rules,
artificial intelligence,
difficulty measure,
correlation coefficient,
random generation
摘要: 给出候选数模式下模仿人工智能求解数独的一系列填数及删减规则,在此基础上提出模仿人工智能的求解算法及数独难度衡量方法。从数独博士5个难度级别中随机抽取各100道题目,采用难度衡量标准重新分级,并将结果与数独博士等级划分标准做相关性检验, 得到Goodman-Kruskal相关系数r=0.82,说明该标准与数独博士的难度划分标准有较强的相关性,并给出随机生成数独题目的算法。通过难度衡量方法与生成算法,可以随机生成5个不同难度的数独谜题。
关键词:
数独,
填数及删减规则,
人工智能,
难度衡量,
相关系数,
随机生成
CLC Number:
TUN Lin-Bei, XIAO Hua-Yong, YANG Zong-Ze. Research of Weight Method with Sudoku Difficulty Level Partition[J]. Computer Engineering, 2012, 38(10): 161-163.
吴林波, 肖华勇, 杨宗泽. 数独难度级别划分的权重法研究[J]. 计算机工程, 2012, 38(10): 161-163.