基于超扩展规则的知识编译方法
发布时间：2018-08-02 14:33
【摘要】：超扩展规则是对扩展规则的扩充,基于超扩展规则能够求得任意两个非互补且不相互蕴含的子句所能扩展出极大项集的交集与差集,并将所得结果以EPCCL理论的形式保存.该文首次提出了扩展反驳方法,是一种新型推理方法,并在该推理方法与知识编译之间建立了联系.基于超扩展规则的性质,该文还提出了两种知识编译算法:求并知识编译算法UKCHER和求差知识编译算法DKCHER,是两种新的知识编译算法.算法UKCHER是目前为止唯一一个可并行的EPCCL理论编译算法,算法DKCHER对于相变点附近的难解问题具有较高的编译效率和编译质量.实验结果表明:UKCHER算法的编译效率和编译质量均优于Lin等人提出的KCER算法;当子句数和变量数的比值较大时,DKCHER算法的编译效率和编译质量是最优的,相比于现有EPCCL理论编译算法,该算法具有较强的竞争力.
[Abstract]:The hyperextended rule is an extension of the extended rule. Based on the hyperextended rule, the intersection and difference set of the maximal itemset can be obtained by any two non-complementary and non-mutual implicative clauses, and the results obtained are preserved in the form of EPCCL theory. In this paper, an extended rebuttal method is proposed for the first time, which is a new reasoning method, and the relationship between the reasoning method and knowledge compilation is established. Based on the properties of superextended rules, this paper also proposes two new knowledge compilation algorithms: the parallel knowledge compilation algorithm (UKCHER) and the differential knowledge compilation algorithm (DKCHER), which are two new knowledge compilation algorithms. The algorithm UKCHER is the only parallel theoretical compilation algorithm for EPCCL up to now. The algorithm DKCHER has high compiling efficiency and high quality for the difficult problem near the phase transition point. The experimental results show that the compiling efficiency and quality of the Lin algorithm are better than that of the KCER algorithm proposed by Lin et al. When the ratio of the number of clauses to the number of variables is large, the compiling efficiency and quality of the DKCHER algorithm are the best, compared with the existing EPCCL theoretical compilation algorithm. The algorithm is competitive.
【作者单位】： 吉林大学计算机科学与技术学院;吉林大学数学学院;符号计算与知识工程教育部重点实验室(吉林大学);
【基金】：国家自然科学基金(61300049) 教育部高等学校博士学科点专项科研基金(20120061120059) 吉林省重点科技攻关项目(20130206052GX) 吉林省青年科研基金项目(20140520069JH,20150520058JH,20150520060JH) 吉林省自然科学基金项目(20150101054JC)资助
【分类号】：TP314
本文编号：2159748
[Abstract]:The hyperextended rule is an extension of the extended rule. Based on the hyperextended rule, the intersection and difference set of the maximal itemset can be obtained by any two non-complementary and non-mutual implicative clauses, and the results obtained are preserved in the form of EPCCL theory. In this paper, an extended rebuttal method is proposed for the first time, which is a new reasoning method, and the relationship between the reasoning method and knowledge compilation is established. Based on the properties of superextended rules, this paper also proposes two new knowledge compilation algorithms: the parallel knowledge compilation algorithm (UKCHER) and the differential knowledge compilation algorithm (DKCHER), which are two new knowledge compilation algorithms. The algorithm UKCHER is the only parallel theoretical compilation algorithm for EPCCL up to now. The algorithm DKCHER has high compiling efficiency and high quality for the difficult problem near the phase transition point. The experimental results show that the compiling efficiency and quality of the Lin algorithm are better than that of the KCER algorithm proposed by Lin et al. When the ratio of the number of clauses to the number of variables is large, the compiling efficiency and quality of the DKCHER algorithm are the best, compared with the existing EPCCL theoretical compilation algorithm. The algorithm is competitive.
【作者单位】： 吉林大学计算机科学与技术学院;吉林大学数学学院;符号计算与知识工程教育部重点实验室(吉林大学);
【基金】：国家自然科学基金(61300049) 教育部高等学校博士学科点专项科研基金(20120061120059) 吉林省重点科技攻关项目(20130206052GX) 吉林省青年科研基金项目(20140520069JH,20150520058JH,20150520060JH) 吉林省自然科学基金项目(20150101054JC)资助
【分类号】：TP314
【相似文献】
相关期刊论文 前7条
1 谷文祥;赵晓威;殷明浩;;知识编译研究[J];计算机科学;2010年07期
2 谷文祥;郭鸿鹤;殷明浩;王金艳;刘日仙;;多值知识编译[J];东北师大学报(自然科学版);2011年04期
3 Tom Bylander;刘逸敏;;一个简单的知识编译模型[J];计算机科学;1992年05期
4 殷明浩;孙吉贵;林海;吴瑕;;可能性扩展规则的推理和知识编译[J];软件学报;2010年11期
5 殷明浩;刘华;谷文祥;;一种基于知识编译的模型计数方法[J];东北师大学报(自然科学版);2010年04期
6 陈永鸿,陈一秀;基于书本知识的多媒体写作系统[J];华侨大学学报(自然科学版);2000年01期
7 ;[J];;年期
相关硕士学位论文 前5条
1 赵晓威;知识编译技术及可能性灵活规划问题研究[D];东北师范大学;2010年
2 刘华;基于知识编译的QBF求解器的研究[D];东北师范大学;2010年
3 牛当当;基于超扩展规则的高效知识编译方法研究[D];吉林大学;2015年
4 逯焱;一种新的基于扩展规则的知识编译方法[D];东北师范大学;2009年
5 邹婷婷;量化布尔范式的近似知识编译方法[D];东北师范大学;2010年
，本文编号：2159748
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2159748.html
