[1]刘　全,等.一种逻辑强化学习的tableau推理方法[J].智能系统学报,2008,3(04):355-360.
　L IU Quan,CU I Zhi-ming,et al.Tableau reason ing method based on logical re inforcement learn ing[J].CAAI Transactions on Intelligent Systems,2008,3(04):355-360.
点击复制

一种逻辑强化学习的tableau推理方法(/HTML)
分享到：

《智能系统学报》[ISSN:1673-4785/CN:23-1538/TP]

卷:
第3卷
期数:
2008年04期
页码:
355-360
栏目:
出版日期:
2008-08-25

文章信息/Info

Title:
Tableau reason ing method based on logical re inforcement learn ing
文章编号:
1673-4785 (2008) 04-0355-06
作者:
刘　全1;  2 ; 崔志明1 ; 高　阳2 ; 陈道蓄2 ; 姚望舒1
1. 苏州大学计算机科学与技术学院,江苏苏州215006;
 2. 南京大学软件新技术国家重点实验室,江苏南京210093
Author(s):
L IU Quan1;  2 ;  CU I Zhi-ming2 ;  GAO Yang1 ;  CHEN Dao-xu1 ;  YAO Wang-shu2
1. School of Cumputer Science of Technology, Soochow University, Suzhou 215006, China;
2. State Key Laboratory for Novel Software Technology, Nanjing University, Nanjing 210093, China
关键词:
逻辑强化学习;  tableau推理
Keywords:
logical reinforcement learning;  tableau reasoning
分类号:
TP301
文献标志码:
A
摘要:
tableau方法是一种具有较强的通用性和适用性的推理方法,但由于函数符号、等词等的限制,使得自动推理具有不确定性. 针对tableau推理中封闭集合构造过程具有盲目性的问题,提出将强化学习用于tableau自动推理的方法. 该方法将tableau推理过程中的逻辑公式与强化学习相结合,产生抽象的状态和活动. 这样一方面可以通过学习方法控制自动推理的推理顺序,形成合理的封闭分枝,减少推理的盲目性;另一方面复杂的推理可以利用简单的推理结果,提高推理的效率.
Abstract:
The tableau method is a reasoning method with high universality and app licability. However, given the restrictions of function symbols and equations, there remains a great deal of uncertainty in automated reasoning. In order to remove blind reasoning in the construction of a closed set for tableau reasoning, a method was developed to introduce reinforcement learning into tableau reasoning. Reinforcement learning was combined with the logical for2 mulae in tableau reasoning to p roduce abstract states and actions. On the one hand, reasoning sequences in auto reasoning can be controlled by the learning method to form reasonable closed branches and reduce the blindness of reasoning. On the other hand, simp le reasoning results can be reused in the comp lex reasoning system to imp rove reasoning efficiency.

参考文献/References:

[ 1 ]史忠植,董明楷,蒋运承,张海俊. 语义Web的逻辑基础 [ J ]. 中国科学(E辑) , 2004, 34 (10) : 112321138.
SH I Zhongzhi, DONGMingkai, J IANG Yuncheng, ZHANG Haijun. A logic foundation for the semantic web [ J ]. Sci2 ence in China ( Series E) , 2004, 34 (10) : 1123 - 1138.
[ 2 ]BLACKBURN P, BOS J. Rep resentation and inference for natural language [ C ] / / CSL I Publications, Crysmann, Berthold, 2005.
[ 3 ]BERTOSS L, SCHW IND C. Analytic tableaux and database repairs [ C ] / /Foundations of Information and Knowledge Systems. Sp ringer LNCS 2284, 2003.
[ 4 ]苏开乐,骆翔宇,吕关锋. 符号化模型检测CTL [ J ]. 计算机学报, 2005, 28 (11) : 179821806.
SU Kaile, LUO Xiangyu, LU¨Guanfeng. Symbolic model checking for CTL [ J ]. Chinese Journal of Computer, 2005, 28 (11) : 179821806.
[ 5 ] PASKEV ICH A. Connection tableauxwith lazy paramodula2 tion[C ] / / IJCAR 2006. Seattle, USA, 2005.
[ 6 ]HORV ITZ E. Machine learning, reasoning, and intelligence in daily life: directions and challenges[C ] / /Proceedings of ICML. Bled, Slovenia, 1999.
[ 7 ]BRYANT C, MUGGLETON S, OL IVER S. Combining in2 ductive logic p rogramming, active learning and robotics to discover the function of genes [ J ]. Electronic Transactions in Artificial Intelligence, 2001, 6 (12) : 1236.
 [ 8 ]CALZONE L, CHABR IER N, FAGES F. Machine learning biomolecular interactions from temporal logic p roperties [C ] / /Proceedings of CMSB 2005. Edinburgh, Scotland, 2005.
[ 9 ] TEEVAN J , HORV ITZ E. Personalizing search via automa2 ted analysis of interests and activities [ C ] / /Proceedings of SIGIR. Salvador, Brazil, 2005: 4492456.
[ 10 ] F ITTINGM. First2order logic and automated theorem p ro2 ving[M ]. New York: Sp ringer2Verlag, 1996.
[ 11 ]高　阳,陈世福,陆　鑫. 强化学习综述[ J ]. 自动化学报, 2004, 30 (1) : 862100.
GAO Yang, CHEN Shifu, LU Xin. Research on reinforce2 ment learning technology: a review [ J ]. Acta Automatica Sinica, 2004, 30 (1) : 862100.
 [ 12 ]OTTERLO M. Reinforcement learning for relationalMDPs [C ] / / Machine Learning Conference of Belgium and the Netherlands (BeNeLearn’04). [ S. l. ] , 2004: 1382145.
 [13 ]刘　全,孙吉贵. 基于Tableau的定理机器证明系统Tab2 leauTAP[J ]. 计算机工程, 2006, 32 (7) : 38245.
 L IU Quan, SUN J igui. Theorem proving system based on Tab2 leauTAP[J ]. Computer Engineering, 2006, 32 (7) : 38 - 45.

备注/Memo

备注/Memo:
收稿日期: 2007-10-22.
基金项目:国家自然科学基金资助项目( 60673092, 60775046) ;教育部重点资助项目(207040) ;中国博士后科研基金资助项目 (20060390919) ; 江苏省高校自然科学基金资助项目 (06KJB520104) ; 江苏省博士后科研基金资助项目 (060211C) ;江苏省现代企业信息化应用支撑软件工程技术研究中心开发项目( SX200804) .
作者简介: 　　
刘　全,男,1969年生,教授,博士后,中国计算机学会高级会员,主要研究方向为智能信息处理、自动推理、机器学习.主持和参与国家级科研项目4项,主持省部级和市(局)级科研项目10多项,获省部级科技进步奖2项,市(局)级科技进步奖8项. 发表学术论文40余篇,其中SCI收录4篇, EI收录20篇. 　　
崔志明,男, 1961年生,教授,博士生导师,中国计算机学会高级会员,主要研究方向为模式识别、Deep Web. 主持国家级及省部级科研项目18 项. 作为项目负责人完成并通过省(部)级以上鉴定的项目有28项,并获国防科工委科技进步一等奖1项、省部级科技进步二等奖2项、三等奖4项、省优秀软件奖三等奖2项;发表学术论文100余篇,其中被SCI、EI收录28篇,出版著(译) 作13部,申请发明专利4项,获软件著作权6项.
高　阳,男, 1972年生,副教授,博士,中国人工智能学会理事,中国机器学习专业委员会常务委员,主要研究方向为强化学习、多Agent系统. 发表学术论文50余篇.
通信作者:刘　全. E-mail: quanliu@suda. edu. cn.
更新日期/Last Update: 2009-05-18