ebook img

Inductive Logic Programming: From Machine Learning to Software Engineering PDF

241 Pages·1995·3.048 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Inductive Logic Programming: From Machine Learning to Software Engineering

InductLiovgei cP rogramming InductLiovgei cP rogramming FromM achinLee arnitnogS oftwaErneg ineering FrancesBceor gadaannod D anieGluen etti TheM IT Press CambridgMea,s sachusetts LondonE,n gland ©1996M assachusIentsttsi touftT ee chnology Allr ighrtess erveNdo. p arto ft hibso okm ay ber eproducienad n yf ormb ya ny electroonrmi ecc hanimceaal ns( includpihnogt ocopyriencgo,r dionrig n,f ormation storaagned r etriewviatlh)o puet rmissiinow nr itifnrgo mt hep ublisher. Thisb ookw ass eti nLl\ TEX by thea uthorasn dw asp rintaendd b oundi nt he UniteSdt ateosfA merica. LibraorfyC ongreCsast aloging-in-PuDbaltiac ation BergadanFor,a nces1c9o6,3 - Inductilvoeg ipcr ogrammifnrgo:m m achinlee arnitnosg o ftware engineer/i Fnrga ncesBceor gadaannod D anielGeu netti. p. cm.-(Logpirco gramming) Includbeisb liograprheifcearle nacnedis n dex. ISBN0 -262-023(9h3c-a:8l kp.a per) 1.L ogipcr ogramminIg.G. u nettDia,n ielIeI.T. i tle. IIIS.e ries. QA 76.63.7B 41 995 95-14956 005.1'1-dc20 CIP Contents SerieFso reword ix Preface xi 1 Introduction 1 I Fundamentals 9 2 ProblemS tatemenatn d Definitions 11 2.1 LogiPcr ogramasn dT heiErx amples 11 2.2 TheI LPP roblem 13 2.3 IncremenStyaslt emasn dQ ueries 22 2.4 IdentifLyoignigPc r ogramisnt heL imit 27 3 Bottom-upM ethods 33 3.1 PlotkiLne'ass Gte neraGle neralization 34 3.2 InverRsees olution 45 3.3 InverIsmep lication 60 4 Top-downM ethods 77 4.1 Shapiro'sI nfMeordeeSnlyc set em 79 4.2 FOIL 85 5 A UnifyinFrga mework 91 5.1 TheorePmr ovinwgi thI nverRsees olution 91 5.2 ExtensioTnoapl- DowMne thodRse visited 99 5.3 Example 103 v VI Contents II ILP withS trongB ias 107 6 InductivBei as 109 6.1 RefinemenOtp erators 111 6.2 ClausTee mplates 115 6.3 DomainT heoriaensd G rammars 118 6.4 Biasi nB ottom-Supy stems 125 6.5 ClausSee ts 129 7 ProgramI nductiowni thQ ueries 137 7.1 TheF ILPS ystem 139 7.2 JustificaotfiE oxnt ensionaanldiP tryo blems 142 7.3 CompletiEnxga mplebse forLee arning 144 7.4 Discussion 147 8 ProgramI nductiowni thouQtu eries 149 8.1 TheI nductiPorno cedure 150 8.2 Example 153 8.3 Propertoifet she I nductiPorno cedure 154 8.4 A SimplifiIemdp lementation 157 8.5 Discussion 163 III SoftwarEen gineeriAnpgp lications 165 9 DevelopmentM,a intenancaen,d Reuse 167 9.1 Introduction 169 9.2 InductLiovgei Pcr ogrammiLnagn guages 171 9.3 The InductiSvoef twaPrreo cess 174 9.4 From InductiLveea rnitnogI nductiPvreo gramming 180 10 Testing 185 10.1I ntroducttoiT oens ting 186 10.2I nductiaonndT estiCnogm pared 187 10.3I nductiTvees tC aseG eneration 189 10.4E xamples 191 11 A CaseS tudy 199 11.1S ynthesiIzinnsge rt 200 11.2T estiIngn sert 209 Contents vii A How toF TP Our Software 217 Bibliography 219 Index 237 SerieFso reword Thel ogipcr ogrammianpgp roatcohc omputiinngv estigtahteue sse o fl ogiacs a programmilnagn guaagned e xplorceosm putatimoondaell bsa seodn c ontroldlee­d duction. Thefi eldo fl ogipcr ogrammihnagss eena tremendogurso wtihn t hel asste veral yearbso,t hi nd eptahn di ns copeT.h igsr owtihsr eflectientd h en umbeorf a rticles, journatlhse,s ebso,o ksw,o rkshopasn,dc onferendceevso tetdot hes ubjecTth.e MIT Pressse riiensl ogipcr ogrammiwnags c reatteoda ccommodatthei dse velop­ menta ndt on urtuirteI. t i sd edicatteotd h ep ublicatoifho ing h-quatleixttyb ooks, monographcso,l lectianodnp sr,o ceediinnlg osg ipcr ogramming. EhudS hapiro TheW eizmanInn stitouftS ec ience RehovoIts,r ael Preface InductLiovgei Pcr ogrammi(nIgL Ph)a se volvferdo mp revioruess earicnMh a chine LearninLgo,g iPcr ogramminagn,dI nductiPvreo graSmy nthesLiisk.er elational MachinLee arninigtd, e alwsi tht hei nductioofcn o nceprtesp resenitnae dl ogical formT.y picaltlhyeo, u tpuotf a nI LPl earniesra seto fH ornc lauseHso.w ever, ILPc annobte j usat n ewn amef orr elatioMnaaclh inLee arninTgh.i sb ooke mpha­ sizetsh ifsa cbty givinsgp eciaatlt entitoo"n P rogramminlge,st"so " Inductive" and" LogicI.L"P t echniqhuaevset hep otenttioas lu pporsto ftwadreev elopment andm aintenanScoem.e o ft hestee chniqudeesv,e lopbeydt hea uthorasr,es tudied inf uldle taainld t heiirm plementatairoemn asd ea vailatbhlreo ugahn onymous ftpT.h eiSro ftwaErneg ineeraipnpgl icatairoetn hse nd iscusisnet dh ec onteoxfta relaticvoemlpyl elxo gipcr ogramH.o wevetrh,i bso oka lshoa st hei mportamnot­ tivatiofop nr ovidiannug p -to-daantdee xtendseudr veoyfI LPa sa researacrhe a. Theb asinco tiontsh,em ostc ommoni nductioopne ratoarnsd,t heb est-known methodasn ds ystemasr ed escribaendda nalyzeCdo.m parewdi tho theerx isting surveytsh,ep resebnoto km ayb em orec ompletaen,di ncludneesw enro tionssu,c h asi nverismep licatnieowne;gr o altsh aatr ei mportafnotrp rogrammiansgs istants, sucha sm ultipplree diclaetaer nianngd;i mportatnotp itchsa atr eo fteonv erlooked, sucha sd eclarabtiiavse. Thisw orki st her esuolftm anyy earosf r esearacnhd,m anyh aved irectolry indirecctolnyt ributtoei dt .W e wisht ot hankL orenzSaa ittaan dA ttilGiioo r­ danaa,s w elals G abrieLloel lfio,rt hee arljyo inwto rkt hahta si nspiroeudrm ore recenpte rspectiavnedfs o,rt heicro ntinusiunpgp orAts. i na lllo ng-terrems earch effortasn,e ssentcioanlt ribuitsir oenp resenbtyed di scussijoonisnw,to rko nr elated topicasn,d c ollaboraitnii notne rnatipornoajle cTthsi.sk indo fs upporitsd uet o manyr esearchers whoa lblce al ninsotthe edr eH.o wevewre, wouldli kteo men­ tiotnh eM achinLee arnianngd A I grouapt t heU niversoiftT yu rint,h eE uropean researchienrvso lviendt heE SPRITI LPp rojecatn,dR yszarMdi chalsaknidh is groupa tG eorgMea sonU niver.s iSttyanM atwina ndC lairNee dellreeca da nd commenteodn a firsdtr afotft heb ookt,h arte sultienad n umbeorf i mprovements andc hangeTsh.e r eseardcehs cribientd h ibso okw asfi nanciaslulpyp ortbeydt he EuropeaUnn ionu,n decro ntra6c0ts2 0( ESPRITB RA ILP)a nd6 156( ESPRIT BRA DRUMS2);b y involvemiennt th eI LPNET PECO networakn di nt heM L­ NET networokfe xcelleanncdeb ;y theI taliCaNnR ,f ort heb ilateprraolj eocnt InductiIvnef ere(nIctea ly-USA). FrancesBceor gadaanndDo a nieGluen etti TurinJ,a nuar1y9 96 InductLiovgei cP rogramming 1 Introduction InducLtoigPvireco gramimssiu nffigc ineentwto lr ye qusitraert thibinosgo w ki th ad efiniAt idoenfi.n itthiahotang s e nersaotmeaedg reeimsfe onutin nd[ 62I]t. stattheaIstn ducLtoigPvireco gramimtsih rneeg s eaarreccaoh v etrhiienn gt ersec­ tioofnM achLienaer nainLndog g Pirco gramTmhiiinsasg b .o luitsk aey tihnaigt t icso ncewrintIehnd d ucatniLdoo ng ic Programming. Indeed: inductive, logic, and programAmlitnhgot.uh ageuh t hoorfti htviyis e dwa test oLb aapcakl wiews islel, rigsekn erlaetasigsnr ge eamnepdnr to paom soer ien formnaottiiIvonend .u ctive LogPirco gram(mcianlIglL eiPdn t hree osftt hbeo o)hk abse ecno ncewrintehd systaenmgdse nemreatlh tohdasat rg ei veexna mpalnepdsr odpurcoeg rIanm s. facatnI, L Ps ystmeamyr eceviavreik oiunsod fis n formaabtoituohtnde e sired prograasim n pubtut,th iisn paultw aiynsc leuxdaemsp olfte hspe r ogrianm­'s put/oubtephuatvT ihooeru .t ptuhtai tps r oduicasle odg pirco gtrhaabmte haves ase xpecotnte hdge i veexna mpolrae tls e,a osnat h igphe rceonftt ahegTmey. p ­ icatlhloeyb ,t aipnreodg rwaimtlshl e bneu seodnn ewe xampnloteg si,vt eotn h e ILsPy stdeumr itnhlgee arpnhiansTgeh .ae b omvaeyb ea no versimplbiufitc ation, foarc ompustceire nititjis us astf t,o romfp rogram synthesTihse from examples. viepwr eseinntt hebidos o ekm phastihfzeae tcsht a etx ampalraeebs s olnuotte ly thoen liyn ptuopt r actIiLmPce atlh oAdnso.t ihmepro rstoaunortfci en formation comferso amp rikonroiw laebdogtueht te a rpgreotg rianmc,l updairntgdi eavlelly­ opesdo ftcwoamrpeo naennpdtr so,p eorftt hineee se dseudb procseudcauhstr hees , numbaenrtd h tey poeft haer gumaennodtf ts h ree turvnaelduT eoss t.r tehsiss observaantadil otsnhof,e a tchtap tr acttiocowalilsl lt ohb aev ee mbienmd odreed compelnevxi ronwmewe onutllsid,tk oel ooaktI LaPs l ogpirocg radme velopment witthhh ee lopfe xampalnedns o,jt u satu tomatic pfrroogemrx aammmpilnegs . Thibso opkr ovaindu eps- toa-ndcdao tmep lseutrevo efIy L Pa sa r eseaarrecah, antdh ecno ncenotnrm aettehsto hdasat re eff ecftoisrvo ef tdweavreel opment.

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.