handl numer rang graph dbase knowledg discoveri graph dbase approach discov cybercrim mine insid threat busi transact process empir comparison graph classif algorithm faster comput direct product kernel graph classif discov anomali multipl norm pattern structur numer data gregress extract featur graph transact regress work propos gregress algorithm set label graph real graph extract complet set subgraph subgraph set correl real userspecifi threshold subgraph set correl subgraph set user threshold gregress incorpor prune mechan base correl subgraph featur output correl subgraph featur prune mechan lead signific speedup experiment result term runtim gregress substanti outperform gspan order magnitud regress model produc approach compar accuraci appli graph dbase anomali detect approach discoveri insid threat learn pattern dynam biolog network substructur analysi metabol pathwai graph dbase relat learn dynam graph dbase relat learn tempor pattern biolog network chang time tempor structur analysi biolog network combin microarrai data graph dbase data mine dynam network empir comparison compress dbase frequenc dbase subgraph mine game dbase simul evalu threat detect seaport environ strateg path plan basi risk time infer graph grammar detect overlap frequent subgraph infer edg replac graph grammar report intern workshop mine graph complex structur mgc learn node replac graph grammar metabol pathwai mine structur anomali graph dbase data paper present graph base approach mine anomali domain anomali consist unexpect entiti relationship alter close resembl anomal behavior introduc algorithm purpos detect anomali type graph algorithm focus specif graph chang minimum descript length principl discov substructur instanc anomal entiti relationship synthet real world data evalu effect algorithm term type anomali algorithm demonstr us examin graph base represent data purpos detect fraud infer edg replac graph grammar discov structur anomali graph dbase data abil mine data repres graph import domain detect structur pattern import area data mine anomali detect fraud work term detect anomali graph base data work statist metric condit entropi measur result limit type anomali specif domain paper present graphbas approach uncov anomali domain anomali consist unexpect entiti relationship deviat resembl nonanomal behavior synthet real world data evalu effect algorithm discov anomali graph base represent data learn supervis graph graph base relat learn gbrl differ logicbas relat learn address induct logic program techniqu differ frequent subgraph discoveri address graph base data mine techniqu learn graph logic present represent issu input data prepar output pattern languag form graph base data mine gbrl focus identifi necessarili frequent pattern graph theoret represent data approach graph base data mine simplif challeng frequenc base approach paper discuss issu futur direct graph base relat learn graph dbase analysi human transfer learn game testb infer node replac graph grammar abstract graph grammar combin relat aspect graph iter recurs aspect string grammar repres import step abil discov knowledg data paper describ approach learn node replac graph grammar approach base previou research frequent isomorph subgraph discoveri extend search frequent subgraph check overlap instanc subgraph input graph subgraph overlap node propos node replac grammar product infer hierarchi product compress portion graph product infer product compress graph valid approach experi gener graph grammar measur system infer origin grammar gener graph describ result real world task chemic mine xml schema induct briefli discuss grammar infer system indic studi extend class learnabl graph grammar anomali detect data repres graph build test bed ai mod base applic graph dbase data mine metabol pathwai graph grammar induct structur data visual program comput program express dimens typic call visual program underli theori visual program languag involv graph grammar graph grammar construct manual construct time consum process demand technic knowledg techniqu automat construct graph grammar part desir induct method infer node replac graph grammar method oper label graph broad applic evalu perform infer graph grammar structur represent correct infer grammar verifi pars graph present train set classif threat multi dsensor secur portal detect anomali cargo graph properti infer node replac recurs graph grammar graph grammar combin relat aspect graph iter recurs aspect string grammar repres import step abil discov knowledg data paper describ approach learn node replac graph grammar approach base previou research frequent isomorph subgraph discoveri extend search frequent subgraph check overlap instanc subgraph input graph subgraph overlap node propos node replac grammar product infer hierarchi product compress portion graph product infer product compress graph valid approach experi gener graph grammar measur system infer origin grammar gener graph describ result real world task chemic mine xml schema induct briefli discuss grammar infer system indic studi extend class learnabl graph grammar intern florida artifici intellig research societi confer report advanc knowledg acquisit represent structur discoveri sequenti dconnect data stream learn architectur autom intellig environ develop technolog system percept perspicaci autom control home workplac environ challeng problem present complet agent architectur learn autom intellig environ discuss develop deploy techniqu util work intellig environ empir evalu approach proven effect reduc inhabit interact structur discoveri sequenti connect data serial partit approach scale graph dbase knowledg discoveri autom hpomdp construct data dmine techniqu intellig environ domain learn exampl singl graph autom intellig smart environ scale ai algorithm larg problem requir algorithm work har respect strength introduc method automat construct hhmm output sequenti data mine algorithm sequenti predict algorithm present theori techniqu demonstr result mavhom intellig environ manag adapt versatil environ graph dbase relat learn applic secur abstract describ approach learn pattern relat data repres graph approach implement subdu system search pattern maxim compress input graph subdu supervis learn unsupervis pattern discoveri cluster appli subdu domain relat homeland secur social network analysi keyword graph relat learn secur editori call amp calendareditori current futur trend featur select extract classif problem manag adapt versatil environ comparison graph dbase logic dbase multi drelat data mine goal paper gener insight differ graph base logic base approach multi relat data mine perform case studi graph base system subdu induct logic program system cprogol identifi kei factor compar graph base logic base multi relat data mine abil discov structur larg concept abil discov semant complic concept abil effect util background knowledg perform experiment comparison subdu cprogol mutagenesi domain artifici gener bongard problem experiment result subdu significantli outperform cprogol discov structur larg multi relat concept observ cprogol learn semant complic concept background knowledg effect subdu agent dbase player dperson entertain dbase real dtime artifici environ structur discoveri sequenti data attribut dvalu select base minimum descript length introduc method attribut select driven minimum descript length principl demonstr viabil approach wisconsin breast cancer data set show work exa mple evalu approach earlier system structur web search engin mdl dbase context dfree graph grammar induct applic algorithm experi test planar graph isomorph identifi inhabit intellig environ graph dbase data mine system goal mavhom smart home project build intellig home environ awar inhabit activ home design provid maximum comfort inhabit minimum cost learn activ inhabit autom activ identifi multipl inhabit present home subdu graph base data mine algorithm discov pattern structur data repres activ pattern inhabit graph subdu inhabit identif introduc multipl class learn version subdu show preliminari result synthet smart home activ data structur web search engin evalu human dconsist behavior real dtime dperson entertain dbase artifici environ mdl dbase context dfree graph grammar induct present algorithm infer context free graph grammar exampl algorithm build earlier system frequent substructur discoveri bias grammar minim descript length grammar featur includ recurs variabl relationship graph dbase data mine system perform web search graph dbase relat learn current futur direct graph base relat learn gbrl differ logicbas relat learn address induct logic program techniqu differ frequent subgraph discoveri address graph base data mine techniqu learn graph logic present represent issu input data prepar output pattern languag form graph base data mine gbrl focus identifi necessarili frequent pattern graph theoret represent data approach graph base data mine simplif challeng frequenc base approach paper discuss issu futur direct graph base relat learn enhanc structur discoveri data mine graphic databas evolutionari program purpos paper develop evolutionari program base system perform data mine databas repres graph import endeavor overemphas data collect nowadai structur natur compos part relat part natur repres graph search capabl evolutionari program util discov concept substructur repeat structur data superior propos techniqu previous develop subdu system cook holder computation constrain beam search space substructur demonstr number data set web domain graph dbase relat concept learn introduc graph base relat concept learn approach implement subduecl system start describ subdu discoveri system graph represent extend perform concept learn task present subduecl graph base concept learner experiment comparison graph dbase relat concept learn induct logic program system compar graph base relat concept learn approach subduecl ilp system foil progol order fair comparison conceptu graph represent conceptu graph standard translat graph logic introduc bia translat process experi type domain show experi artifici domain describ subduecl perform conceptu graph represent experi flat relat domain result comparison show subduecl system competit ilp system flat relat domain structur web search graph dbase discoveri system access inform interest internet present challeng scientist analyst desir inform structur natur goal design structur search engin hyperlink structur web addit textual inform search site interest design structur search engin subdu graph base discoveri tool tool call websubdu enhanc wordnet featur engin search synonym term search engin retriev site structur form graph base user queri demonstr approach number structur web queri client dserver interact tool integr artifici intellig curriculum goal project increas student interest artifici intellig promot learn topic compris subject describ develop web base multimedia deliveri method accomplish goal outlin plan continu develop tool highlight materi integr simul environ student develop test ai algorithm dynam uncertain visual environ graph dbase concept learn introduct concept learn machin learn techniqu learn process driven provid posit neg exampl learner exampl learner build hypothesi concept describ posit exampl exclud neg exampl induct logic program ilp system successfulli concept learner exampl foil progol main engin system base order logic research introduc graph base relat concept learn system call subduecl experi shown competit ilp system type domain subduecl extens made subdu system unsupervis graph base learner extens made subdu provid concept learn capabl relat abil accept input posit neg exampl criteria cover stori structur web search graph dbase discoveri system graph dbase hierarch conceptu cluster hierarch conceptu cluster proven data mine techniqu graph base represent structur inform shown success knowledg discoveri subdu substructur discoveri system advantag approach paper present subdu focu cluster capabl exampl illustr valid approach structur unstructur domain compar subdu earlier cluster algorithm graph dbase hierarch conceptu cluster hierarch conceptu cluster proven data mine techniqu graph base represent structur inform shown success knowledg discoveri subdu substructur discoveri system advantag approach paper present subdu focu cluster capabl exampl illustr valid approach structur unstructur domain compar subdu earlier cluster algorithm graph base concept learn introduct concept learn machin learn techniqu learn process driven provid posit neg exampl learner exampl learner build hypothesi concept describ posit exampl exclud neg exampl induct logic program ilp system successfulli concept learner exampl foil progol main engin system base order logic research introduc graph base relat concept learn system call subduecl experi shown competit ilp system type domain subduecl extens made subdu system unsupervis graph base learner extens made subdu provid concept learn capabl relat abil accept input posit neg exampl criteria graph dbase hierarch conceptu cluster structur databas introduct cluster analysi studi develop area wide varieti applic purpos appli cluster databas gain understand data case reveal hierarch topolog work extend subdu structur knowledg discoveri system cluster function past work relat increment approach call cobweb fisher extens labyrinth thompson amp langlei repres structur object probabilist model conceptu cluster subdu subdu holder cook knowledg discoveri system deal structur data work graph represent includ vertex edg label direct undirect edg object data map vertic relationship attribut map edg subdu discoveri algorithm discov interest repetit subst discov structur pattern telecommun data increas amount complex data collect urgent creat autom techniqu mine data data gener store telecom compani overwhelm scientist abil manual discov pattern data data structur natur compos part relat part linear attribut base algorithm captur intricaci data exist develop scalabl tool analyz discov concept structur databas introduct technolog law chang telecommun industri blind rate method mine data need understand control amount collect data exce abil manual search interpret pattern data improv discoveri knowledg larg databas equipm structur knowledg discoveri analyz earthquak activ subdu structur discoveri system data mine tool studi amp quot orizaba fault amp quot locat mexico part research project geologist dr burk burkart analyz inform earthquak databas discov earthquak activ area relat fault experi sampl data heurist guid subdu substructur discoveri process ad spatio tempor inform background knowledg result show subdu successfulli data mine tool real world domain graph dbase hierarch conceptu cluster hierarch conceptu cluster proven data mine techniqu graph base represent structur inform shown success knowledg discoveri subdu substructur discoveri system advantag approach paper present subdu focu cluster capabl exampl illustr valid approach structur unstructur domain compar subdu earlier cluster algorithm graph dbase data mine increas amount complex todai data urgent acceler data mine larg databas data structur natur compos part relat part exist develop scalabl tool analyz discov concept structur databas goal research provid system perform data mine databas repres graph paper demonstr subdu system perform kei techniqu data mine unsupervis pattern discoveri supervis concept learn exampl applic subdu system databas repres cad circuit protein molecul databas uc irvin repositori present demonstr scalabl eectiv graph base system introduct larg amount data collect todai quickli overwhelm researc structur knowledg discoveri chemic spatio dtempor databas appli subdu substructur discoveri system chemic toxic domain increas number chemic compound ad year accompani similar growth abil analyz classifi compound problem prevent cancer caus chemic great scientif humanitarian ai discoveri tool predict chemic toxic investig basic idea work obtain structur activ represent sar srinivasan al relat molecular structur cancer activ data obtain nation toxicolog program conduct nation institut environment health scienc nieh gener approach automat discov repetit substructur dataset outlin research relev sar identifi subdu substructur discoveri system discov commonli occur substructur set compound substructur subdu applic knowledg discoveri molecular biologi identifi structur regular protein discov concept structur data introduct autom knowledg discoveri essenti extract inform databas complex databas move simplist attribut represent assum recent knowledg discoveri system structur data represent reflect relationship object discov concept structur data requir identif common substructur data motiv process find substructur capabl compress data abstract instanc substructur identifi conceptu interest substructur enhanc interpret data substructur discoveri process identifi concept describ interest repetit substructur structur data discov substructur concept simplifi data replac instanc substructur pointer newli discov concept experiment ddriven knowledg acquisit plan knowledg engin plan expens result knowledg imperfect autonom learn plan oper definit environment feedback learn system wiser explor instanti liter space breadth search techniqu node search tree repres state uniqu subset instanti liter space state root node call seed state wiser gener seed state util imperfect expert knowledg wiser experi oper node posit state oper successfulli execut constitut initi precondit oper analyz number requir experi function number miss precondit seed state introduc naiv domain assumpt test subset exponenti state space breadth rst search expens wiser introduc search techniqu reorder liter knowledg discoveri molecular biologi identifi structur regular protein recent year explos amount molecular biologi inform obtain deposit databas identifi interpret interest pattern massiv amount inform essenti compon direct molecular biologi research goal research discov structur regular protein sequenc appli subdu discoveri system databas found brookhaven protein data bank paper discuss issu relev applic includ data prepar represent report result appli subdu class protein structur discuss potenti signic result studi protein support nsf grant ili thecb grant introduct topic nding biolog meaning pattern sequenc secondari tertiari threedimension structur protein mac exploit parallel structur scientif discoveri system improv scalabl coupl complementari knowledg discoveri system exploit parallel knowledg discoveri system improv scalabl larg amount data collect todai quickli overwhelm research abil interpret data discov interest pattern knowledg discoveri data mine approach hold potenti autom interpret process approach frequent util computation expens algorithm scienti discoveri system focu util richer data represent regard scalabl research outlin gener approach scale kdd system parallel distribut resourc appli suggest strategi subdu knowledg discoveri system subdu discov interest repetit concept graph base databas varieti domain requir substanti amount process time experi demonstr scalabl parallel version subdu system perform cad circuit databas arti cialli gener databas potenti achiev obstacl discuss improv scalabl scientif discoveri system exploit parallel larg amount data collect todai quickli overwhelm research abil interpret data discov interest pattern knowledg discoveri data mine approach hold potenti autom interpret process approach frequent util computation expens algorithm research outlin gener approach scale kdd system parallel distribut resourc appli suggest strategi subdu knowledg discoveri system subdu discov interest repetit concept graph base databas varieti domain requir substanti amount process time experi demonstr scalabl parallel version subdu system perform cad circuit databas artifici gener databas potenti achiev obstacl discuss introduct barrier integr scientif discoveri method emprir studi domain knowledg benefit substructur discoveri user dcenter system decomposit dbase requir cluster scalabl discoveri inform structur concept domain knowledg discov repetit function substructur larg structur databas improv abil interpret compress data scientist work databas area expertis search predetermin type structur structur exhibit characterist specif domain paper present method guid discoveri process domain specif knowledg paper subdu discoveri system evalu benefit domain knowledg guid discoveri process result show domain specif knowledg improv search substructur domain lead greater compress data empir theoret result scalabl algorithm increasingli larg structur databas keyword data mine minimum descript length principl data compress inexact graph match domain knowledg scalabl support nasa gran decis dtheoret cooper sensor plan paper describ decis theoret approach cooper sensor plan multipl autonom vehicl execut militari mission autonom vehicl applic intellig cooper reason select optim vehicl view locat select optim camera pan tilt angl mission decis made maxim inform gain sensor maintain vehicl stealth mission involv multipl vehicl cooper balanc work load increas inform gain paper present theoret foundat cooper sensor plan research describ applic techniqu arpa unman ground vehicl program introduct tradition research multi agent plan research imag understand pursu independ arpa unman ground vehicl ugv program pull intermedi decis tree intermedi decis tree subtre full unprun decis tree gener breadth order extens empir investig evalu classif error intermedi decis tree compar perform full prune tree empir result gener databas uci machin learn databas repositori result show attempt minim error prune tree produc intermedi tree perform significantli databas result question effect decis tree prune strategi suggest consider full tree intermedi result reveal specif properti satisfi databas intermedi full tree perform relationship improv guidelin select induct strategi base domain properti introduct numer decis tree substuctur discoveri subdu system analyz benefit domain knowledg substructur discoveri discov repetit interest function substructur structur databas improv abil interpret compress data scientist work databas area expertis search predetermin type structur structur exhibit characterist specic domain paper present method guid discoveri process domain specic knowledg paper subdu discoveri system evalu bene ts domain knowledg domain knowledg incorpor subdu singl gener methodolog guid discoveri process result show domain specic knowledg improv search substructur domain lead greater compress data illustr bene ts exampl experi domain comput program comput aid design circuit seri artici gener domain knowledg discoveri structur data discov repetit substructur structur databas improv abil interpret compress data paper describ subdu system domain independ domain depend heurist find interest repetit structur structur data substructur discoveri techniqu discov fuzzi concept compress data descript formul hierarch substructur definit exampl domain scene analysi chemic compound analysi comput aid design program analysi demonstr benefit discoveri techniqu introduct larg amount data collect todai quickli overwhelm research abil interpret data discov interest pattern data respons problem number research develop techniqu discov concept databas techniqu work data express structur attribut empir approach solv gener util problem speedup learn recent approach avoid util problem speedup learn eventu degrad perform due increas amount learn problem solver control knowledg reli sophist util measur signific number train problem accur estim util control knowledg empir result present simpl select strategi retain control rule deriv train problem solut quickli defin effici set control knowledg train problem simpl select strategi low cost altern intens approach improv speed problem solver experiment illustr exist minimum repres cost learn curv reach train exampl stress control amount learn knowledg oppos knowledg attempt made relat domain ch substuctur discoveri subdu system substructur discoveri minimum descript length background knowledg abil identifi interest repetit substructur essenti compon discov knowledg structur data describ version subdu substructur discoveri system base minimum descript length principl subdu system discov substructur compress origin data repres structur concept data replac previous discov substructur data multipl pass subdu produc hierarch descript structur regular data subdu computation bound inexact graph match identifi similar ident instanc substructur find approxim measur close substructur comput constraint addit minimum descript length principl background knowledg subdu guid search substructur experi varieti domain demonstr subdu substructur discoveri minimum descript length background knowledg abil identifi interest repetit substructur essenti compon discov knowledg structur data describ version subdu substructur discoveri system base minimum descript length principl subdu system discov substructur compress origin data repres structur concept data replac previous discov substructur data multipl pass subdu produc hierarch descript structur regular data subdu computation bound inexact graph match identifi similar ident instanc substructur find approxim measur close substructur comput constraint addit minimum descript length principl background knowledg subdu guid search substructur experi varieti domain demonstr subdu discoveri inexact concept structur data introduct autom knowledg discoveri essenti extract inform databas complex databas move simplist attribut represent assum recent knowledg discoveri system structur data represent reflect relationship object discov concept structur data requir identif common substructur data motiv process find substructur capabl compress data abstract instanc substructur identifi conceptu interest substructur enhanc interpret data substructur discoveri process identifi concept describ interest repetit substructur structur data discov substructur concept simplifi data replac instanc substructur pointer newli discov concept empir analysi gener util problem machin learn overfit problem induct learn util problem speedup learn describ common behavior machin learn method eventu degrad perform due increas amount learn knowledg plot perform chang knowledg execut learn method perform respons reveal similar curv method perform respons gener increas singl peak gradual decreas perform similar perform respons suggest model relat perform amount learn knowledg paper empir evid exist gener model plot perform respons learn program formal model perform respons discuss model control amount learn avoid degrad perform introduct machin learn method acquir increas intellig machin learn algorithm fuzzi substructur discoveri paper describ method discov substructur data fuzzi graph match previou implement subdu system discov substructur base psycholog motiv criteria cognit save compact connect coverag instanc data match discov substructur describ implement subdu emploi fuzzi graph match discov substructur occur data form fuzzi substructur discoveri formul fuzzi concept compress data descript discov interest structur data found pure form slightli convolut form exampl domain scene analysi chemic compound analysi demonstr fuzzi discoveri techniqu introduct substructur discoveri process identifi concept describ interest repetit chunk st applic machin learn mainten knowledg dbase perform develop expert system involv knowledg acquisit support appli machin learn techniqu paper present basic idea decis tree induct process optim develop domain model electrochem machin ecm discuss decis tree induct build refin knowledg base process idea develop intellig supervisori system learn compon imafo intellig manufactur foreman implement briefli introduc result appli imafo analyz data form ecm process present domain model process electrochem machin built initi inform result decis tree induct optim model process refin knowledg base shown exampl demonstr rule includ knowledg gener util problem machin learn overfit problem induct learn util problem speedup learn describ common behavior machin learn method eventu degrad perform due increas amount learn knowledg plot perform chang knowledg execut learn method perform respons reveal similar curv method perform respons gener increas singl peak gradual decreas perform similar perform respons suggest model relat perform amount learn knowledg paper empir evid exist gener model plot perform respons learn program formal model perform respons discuss model control amount learn avoid degrad perform introduct machin learn method acquir increas acceler learn connect machin complex machin learn techniqu improv transform iter compon parallel equival parallel consid theori implement perform exist parallel machin parallel architectur connect machin platform implement evalu parallel learn techniqu architectur connect machin limit languag interfac constrain implement learn program connect machin implement learn program perceptron aq comput complex compar sequenti version actual run connect machin techniqu parallel id analyz advantag disadvantag parallel implement connect machin discuss context machin learn empir substructur discoveri intellig machin learn algorithm 