文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > Institute for Informatics, Academy of Sciences of Russia,
侵权投诉

Institute for Informatics, Academy of Sciences of Russia,

If to compare the existing general purpose theorem provers, which efficiency is more than modest, and human way of reasoning, which is incomparably more efficient, one can see some principal distinction in knowledge organization in both cases. We give a no

MinimizingEntropyofKnowledgeRepresentaion

AnatolSlissenko1

Dept.ofInformatics,UniversityParis-12,France

and

InstituteforInformatics,AcademyofSciencesofRussia,

St-Petersburg,Russia

Abstract

Iftocomparetheexistinggeneralpurposetheoremprovers,whiche ciencyismorethanmodest,andhumanwayofreasoning,whichisincomparablymoree cient,onecanseesomeprincipaldistinctioninknowledgeorganizationinbothcases.Wegiveanotionofentropywhichpermitstosetupathesisthatane cientknowledgeorganizationadherestotheprincipleofminimizingtheentropy.Withinthesameframeworkonecande nevariousquantitativecharacteristicsofsystemsofknowledge.Apracticalconsequenceofthispointofviewisstrategy-orientedarchitectureoftheoremprovers.Wementionanexperimentalimplementationofsuchanarchitecturethatdemonstratedpromisingresults.

0.Introduction

Thetopicunderconsiderationismotivatedbyine ciencyofautomatictheorem-proversclearlyunderstoodalreadyinthe60ies.Sincethentherewerenoprincipalprogressinthedevelop-mentofarchitecturesoftheoremprovers,thoughthemodernones(onecan ndmanyreferenciesatdoc.guandang.net)areincomparablymorepowerfulthenthosemadebypioneers[AR83a,AR83b].Buttherelativepowerofmoderntheoremproversisaconsequenceofdevelopmentofcomputersandprogrammingbutnotoffundamentalideas.Thestrategyusedeveninmoderntheoremproversrepresentgeneralsyntacticobservationsthatpracti-callydoplaynoroleinnontrivialproofs.Theyareequallyapplicabletoanydomain.Ontheotherhand,tolearntoreasoninaparticulardomaintakesalotoftime,andmoreover,thestructureofknowledgeofanyparticulardomaindevelops,aswellasmethodsofsearchforproofs,asaresultofconsiderableintellectualinvestments.E.g.compareatwopageformalisationofcalculusandahandbookonthesamesubject.Theoreticalpropertiesofmoreorlesspowerfullogicsthatrep-resentourknowledge,inparticulartheiruniversality,signifythatnogeneralpurposeproofsearchcanbee cient.Theuniversalitymeansthatanyalgorithmicdi cultycanbemodeledintheselogics;thereareevenindividualprovablepropositionsthatarehardtohandle(cf.[Sli75]).Theseconsiderationsmeanthattheonlywaytoachievethee ciencyofproofsearchistoorganizethebasicknowledgeofconcretedomainsandtheknowledgeaboutproofsearchinthesedomainsinane cientway.Butconstructingsuchaknowledgerepresentationneedsadequatetoolstorestructurethisknowledge.Wearegoingtodiscussonetheoreticalpointofviewandonepracticalapproachtodevelopsuchtools.

Inthepresentcommunicationwebrie ydevelopasystemofnotionsthatpermitstogivequantitativecharacteristicsofsystemsofknowledgerepresentation.Thelatterarerepresentedasinferencesystemsde nedinSection1.TheninSection2weintroduceproofcomplexitythatcanservetode nevariousmeasuresofqualityofsystemsofknowledgerepresentation(see[Sli91]).Herewespeakonlyaboutentropy.ThelastSection3touchesastrategybasedtheoremproving.Address:Dept.ofInformatics,UniversityParis-12,61Av.duG´en.deGaulle,94010,Cr´eteil,France.E-mail:slissenko@univ-paris12.fr1

第1页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email