summa subgraph match massiv graph propag bug fix fast subgraph match dessin mine dens subgraph pattern singl graph mine high ddimension data sapper subgraph index approxim match larg graph emerg applic comput biologi softwar engin techniqu social network data form graph locat occurr queri graph larg databas graph import research topic due exist nois miss edg larg databas graph investig problem approxim subgraph index find occurr queri graph larg databas graph miss edg sapper method propos solv problem util hybrid neighborhood unit structur index sapper take advantag pre gener random span tree carefulli design graph enumer order real synthet data set emploi demonstr effici scalabl approxim subgraph index method gaddi distanc index base subgraph match biolog network ring integr method frequent repres subgraph mine arc dmotif discov correl motif unalign biolog sequenc permu dpattern discoveri mutabl permut pattern proxim constraint ram random approxim graph mine gene pattern mine algorithm interchang gene set prokaryot microarrai data mine landmark gene dguid cluster discov neglect condit softwar mine depend graph introduct special issu bioinformat monkei approxim graph mine base span tree treepi graph index method data mine bioinformat biokdd pathfind mine signal transduct pathwai segment protein dprotein interact network data mine bioinformat bioinformat studi applic comput method life scienc data present enjoi surg interest main reason public near complet sequenc human genom anticip knowledg deriv process great impact modern medicin pharmaceut industri expect util knowledg drug design interest bioinformat structur data trajpattern mine sequenti pattern imprecis trajectori mobil object annot protein mine protein interact network doi bioinformat btl biolog text retriev system base background knowledg user feedback arc aggreg relat column score scheme align sequenc motiv biologist frequent align multipl biolog sequenc determin consensu sequenc search predomin residu conserv region determin conserv region align import activ protein sequenc hundr residu longer difficult distinguish biolog import conserv region motif domain wide tool logo al confind entropi base method fail highlight region comput tool highlight biolog import region accur highli desir result paper present score scheme arc aggreg relat column score align biolog sequenc arc method consid tradit charact similar measur column correl extens experiment evalu prosit pattern arc highlight motif region accuraci top peak avail sourc code effici classif multipl databas relat crossmin approach biokdd data mine bioinformat knowledg guid analysi microarrai data biokdd data mine bioinformat gene team relax proxim constraint function relat gene evolv due strong select pressur evolut expect present multipl genom physic proxim gene gene team concept discov function relat gene multipl genom gene set preserv physic proxim paper gener gene team model gene cluster physic cluster form multipl genom case relax constraint propos hybrid pattern model combin set sequenti pattern model model search gene cluster physic proxim constraint model implement test genom replicon result analyz show us model analysi gene cluster belong subtili coli demonstr model predict experiment verifi operon function relat cluster program fast provid sevic web integr heterogen microarrai data sourc correl signatur abstract microarrai latest breakthrough experiment molecular biologi thousand research group gener ten thousand microarrai gene express profil base tissu speci condit combin vast amount microarrai data set import challeng problem paper introduc correl signatur method coher interpret integr microarrai data dispar sourc propos algorithm build gene row tabl correl signatur captur system wide depend exist gene gene tabl compar signatur tabl analysi valid framework experiment studi real microarrai data set result suggest approach viabl solut microarrai data integr analysi problem data mine techniqu microarrai dataset find repres set massiv data inform ag data pervas applic data explos signific phenomenon massiv data volum pose challeng human user comput project propos model identifi repres set larg databas repres set special subset origin dataset main characterist significantli smaller size compar origin dataset captur inform origin dataset compar subset size low redund repres informationtheoret measur mutual inform rel entropi measur repres repres set design greedi algorithm present heurist algorithm deliv perform run experi real dataset evalu effect repres set term coverag accuraci experi show repres set attain expect characterist captur inform effici search relat object relat databas discov knowledg retriev inform relat databas user find object relat sourc object main challeng build ectiv object search system huge amount object databas larg number di erent relationship object paper introduc ross cient accur relat object search system ross accept complex queri enabl user relationship object measur relationship join path ross consid di erent semant di erent join combin select length join path measur strength approach find join path relat convert databas schema graph shortest path graph correspond join path databas ross stream base system architectur handl complex queri logic oper find relat object user request comprehens experi conduct show high scalabl ectiv ross survei biodata analysi data mine perspect mine high ddimension data make svm scalabl larg data set hierarch cluster index support vector machin svm promis method classif regress analysi due solid mathemat foundat includ desir properti margin maxim linear classif kernel promin properti svm chosen larg scale data mine problem train complex highli depend data set size unlik tradit pattern recognit machin learn real world data mine applic involv huge num ber data record expens perform multipl scan entir data set infeas put data set memori paper present method cluster base svm cb svm maxim svm perform larg data set limit amount resourc memori cb svm appli hierarch micro cluster algorithm scan entir data set provid svm high qualiti sampl sampl carri statist summari data maxim benefit learn analys show train complex cb svm quadrat depend number support vector entir data set experi synthet real world data set show cb svm highli scalabl larg data set accur term classif system user dguid biolog literatur search engin guest editor introduct special issu mine biolog data improv biclust method analyz gene express profil biclust gene express data tendenc advent dna microarrai technolog revolution experiment studi gene express cluster popular approach analyz gene express data proven success applic work focus discov subset gene exhibit similar express pattern subset condit gene express matrix specif order preserv cluster op cluster subset gene induc similar linear order subset condit pioneer work opsm model enforc strict order share gene cluster includ model special case model robust opsm similarli express condit allow form order equival group restrict order group guid model design implement determinist algorithm opc tree discov op cluster experiment studi real dataset demonstr effect algorithm applic tissu classif cell cycl identif addit larg percentag op cluster exhibit signific enrich function categori impli op cluster carri signific biolog relev subspac cluster microarrai data analysi multipl criteria signific assess gene ontolog friendli biclust express profil sound cluster analysi gene express profil gene function predict base hypothesi gene similar express profil impli strong correl function biolog activ gene ontolog accept standard organ gene function categori gene function categori sophist relationship part overlap cluster algorithm gener gene cluster relationship natur reflect gene function categori hierarchi failur resembl relationship reduc confid cluster gene function predict paper present cluster subspac cluster microarrai data analysi multipl criteria signific assess crossmin effici classif multipl databas relat todai structur data store relat databas databas consist multipl relat link conceptu entiti relationship link design relat databas schema multi relat classif wide disciplin financi decis make medic research geograph applic classif approach work singl flat data relat difficult convert multipl relat singl flat relat introduc huge undesir univers relat lose essenti inform previou work induct logic program approach recent relat mine proven effect high accuraci multi relat classif suffer poor scalabl number relat number attribut databas crossmin effici classif multipl databas relat todai structur data store relat databas databas consist multipl relat link conceptu entiti relationship link design relat databas schema multi relat classif wide disciplin financi decis make medic research geograph applic classif approach work singl flat data relat difficult convert multipl relat singl flat relat introduc huge undesir univers relat lose essenti inform previou work induct logic program approach recent relat mine proven effect high accuraci multi relat classif suffer poor scalabl number relat number attribut databas agil gener approach detect transit evolv data stream applic commerc system diagnosi telecommun servic data arriv stream high speed common underli process gener stream chang time result fundament evolut respons extern stimulu detect challeng problem great practic import volum stream exce main memori access data stream typic perform linear scan ascend order indic record paper propos approach agil monitor stream data detect distinguish transit underli process agil advantag tradit hidden markov model agil requir scan data spin mine maxim frequent subgraph graph databas fundament challeng mine recur subgraph semi structur data set overwhelm abund pattern larg graph databas total number frequent subgraph larg full enumer reason comput resourc paper propos algorithm mine maxim frequent subgraph subgraph part frequent subgraph exponenti decreas size output set case experi practic data set mine maxim frequent subgraph reduc total number mine pattern order magnitud method mine frequent tree gener graph databas reconstruct maxim subgraph mine tree chemic structur benchmark set synthet graph data set demonstr addit decreas output size algorithm achiev fold speed current state art subgraph mine algorithm cluster move object framework ontolog ddriven subspac cluster tradit cluster descript task seek identifi homogen group object base valu attribut domain knowledg justifi cluster cluster algorithm domain knowledg consider paper domain knowledg repres hierarch ontolog develop framework directli incorpor domain knowledg cluster process yield set cluster strong ontolog implic cluster process ontolog inform util effici prune exponenti search space subspac cluster algorithm algorithm gener automat interpret cluster result map natur hierarch organ subspac cluster signific categor enrich ontolog hierarchi experi set gene express data gene ontolog demonstr prune techniqu driven ontolog significantli improv cluster perform minim degrad cluster qualiti hierarch organ gene cluster hierarchi gene ontolog successfulli captur continu dnearest neighbor search move object paper describ method continu monitor nearest neighbor object mobil environ monitor nearest neighbor choos monitor nearest neighbor condit knn chang neighbor addit origin space move object transform time distanc td space object repres curv beach line algorithm develop monitor chang neighbor enabl maintain knn increment extens empir studi show beach line algorithm outperform effici exist algorithm wide margin total number object larg bass approxim search larg string databas paper studi problem build index structur larg string databas effici support type string match necess map substr numer space string tree mr index restrict memori practic suffix tree suffix arrai goal propos index scheme bass tree effici support gener approxim substr match term symbol substitut misalign sublinear time larg string databas kei idea design posit string group recurs fulli balanc tree similar subsequ segment start posit node label regular express describ common substr index subtre search properli direct portion databas high potenti match quickli bass tree place wild card queri pattern handl seamless addit search long pattern decompos seri search short segment process join result demonstr experi potenti perform improv brought bass tree order magnitud altern method mine surpris period pattern paper focu mine surpris period pattern sequenc event applic comput biologi infrequ pattern consid signific actual occurr frequenc exce prior expect larg margin tradit metric support necessarili ideal model measur kind surpris pattern treat pattern equal sens occurr carri weight assess signific pattern probabl occurr suitabl measur inform introduc natur degre surpris occurr pattern continu monoton decreas function probabl occurr pattern vastli occurr probabl handl seamlessli accumul degre surpris repetit pattern concept inform gain propos measur degre surpris pattern data sequenc bound inform gain properti identifi tackl predica caus violat downward closur properti inform gain measur turn effici solut problem empir test demonstr effici us propos model war weight associ rule item intens discov high dorder period pattern discoveri period pattern time seri data activ research area applic pattern hierarch natur higher level pattern consist repetit lower level pattern presenc nois prevent higher level pattern recogn sens portion data sequenc support high level pattern di erent layout occurr basic symbol exist common represent term raw symbol combin high level pattern express previou model defin raw symbol symbol combin properli recogn exist method paper propos model meta pattern captur high level pattern exibl model number potenti meta pattern larg substanti diculti li identifi proper pattern candid priori properti provid suffici prune power properti compon locat properti identifi conduct candid gener ecient comput base mine algorithm develop appli algorithm real synthet sequenc interest pattern discov discov compact highli discrimin featur featur combin drug activ support vector machin nowadai high throughput experiment techniqu make feasibl examin collect massiv data molecular level data typic map high dimension featur space carri rich inform function chemic biolog entiti infer valuabl knowledg purpos classif predict typic small number featur featur combin plai determin role function discrimin identif featur featur combin great import paper studi problem discov compact highli discrimin featur featur combin rich featur collect emploi support vector machin classif mean aim find compact featur combin compar previou method featur select identifi featur sole base individu role classif method identifi minim featur combin ultim determin role systemat fashion experiment studi drug activ data show method discov descriptor necessarili signific individu signific collect enhanc biclust express data molecular biologi provid power tool express pattern thousand gene monitor simultan produc huge amount valuabl data concept biclust introduc cheng church captur coher subset gene subset condit set heurist algorithm design find biclust set biclust consist iter mask null valu discov biclust coars fine node delet node addit inclus invert data heurist inevit suffer drawback mask null valu discov biclust random number result phenomenon random interfer turn impact discoveri high qualiti biclust address issu acceler biclust process gener model biclust incorpor null valu propos probabilist algorithm floc discov set possibl overlap biclust simultan algorithm easili extend support addit featur suit requir virtual cost experiment studi yeast gene express data show floc algorithm offer substanti improv previous propos algorithm discov compact highli discrimin featur featur combin drug activ support vector machin nowadai high throughput experiment techniqu make feasibl examin collect massiv data molecular level data typic map high dimension featur space carri rich inform function chemic biolog entiti infer valuabl knowledg purpos classif predict typic small number featur featur combin plai determin role function discrimin identif featur featur combin great import paper studi problem discov compact highli discrimin featur featur combin rich featur collect emploi support vector machin classif mean aim find compact featur combin compar previou method featur select identifi featur sole base individu role classif method identifi minim featur combin ultim determin role systemat fashion experiment studi drug activ data show method discov descriptor necessarili signific individu signific collect deliv multimedia stream flexibl qo multicast dag cluseq effici effect sequenc cluster analyz sequenc data increasingli import recent area biolog sequenc text document web access log paper investig problem cluster sequenc base structur featur wide recogn techniqu cluster proven detect unknown object categori reveal hidden correl object difficulti prevent cluster perform extens sequenc data categor domain lack effect effici similar measur propos model cluseq sequenc cluster explor signific statist properti possess sequenc condit probabl distribut cpd symbol preced segment deriv character sequenc behavior support similar measur variat suffix tree probabilist suffix tree emploi organ signific portion cpd concis algorithm devis effici discov cluster high qualiti automat adjust number cluster optim rang uniqu combin success cluster gener cluster consolid perform cluseq demonstr extens experi real synthet sequenc databas dynam cluster evolv stream singl pass classifi larg data set svm hierarch cluster carpent find close pattern long biolog dataset characterist dataset typic larg number column small number row gene express dataset column row stamp discoveri statist import pattern repeat long sequenti data paper focu mine period pattern allow degre imperfect form random replac perfect period pattern infomin propos metric gener inform gain identifi pattern event vastli occurr frequenc adjust deviat pattern penalti allow gap pattern occurr locat repeat dna sequenc paper present effect mine algorithm stamp simultan mine signific pattern subsequ model gener inform gain mine asynchron period pattern time seri data periodici detect time seri data challeng problem great import applic recent progress select topic databas research report young chines research work unit state automat cluster protein sequenc analyz protein sequenc data increasingli import recent previou work area focus build classif model paper investig problem automat cluster unlabel protein sequenc wide recogn techniqu statist comput scienc cluster proven detect unknown object categori reveal hidden correl object difficulti prevent cluster perform directli protein sequenc lack effect similar measur comput effici propos model protein sequenc cluster explor signific statist properti possess sequenc concept imprecis probabl introduc origin probabilist suffix tree monitor converg empir measur guid cluster process demonstr propos method successfulli discov meaning famili necess learn model famili pre label train data acceler approxim subsequ search larg protein sequenc databas framework effici effect sequenc clusterin delta dcluster captur subspac correl larg data set cluster activ research area great practic import recent year previou cluster model focus group object similar valu set dimens subspac cluster assum object dimens biclust exist cluster model adequ captur coher exhibit object strong coher exist set object subset attribut valu attribut attribut valu fulli common applic includ bio informat analysi collabor filter analysi data incomplet subject bias bio informat biclust model recent propos captur coher subset attribut introduc gener model refer ffi cluster model captur coher exhibit subset object subset attribut allow absent attribut valu move base algorithm floc devis effici produc optim cluster result ffi cluster model take biclust model special case floc algorithm perform superior biclust algorithm demonstr correct effici ffi cluster model floc algorithm number real synthet data set infomin mine partial period pattern gap penalti paper focu mine period pattern allow degre imperfect form random replac perfect period pattern inform gain propos identifi pattern event vastli occurr frequenc adjust deviat pattern penalti exist gap pattern occurr applic bio informat import identifi subsequ pattern repeat perfectli perfectli solut extend inform gain measur includ penalti gap pattern occurr call measur gener inform gain find subsequ pattern gener inform gain high locat repeat dna sequenc paper develop effect mine algorithm infomin simultan mine signific pattern subsequ effici filter larg dataseta user dcentric paradigm cluster pattern similar larg data set cluster process group set object class similar object definit similar vari cluster model model concept similar base distanc euclidean distanc cosin distanc word similar object requir close valu set dimens paper explor gener type similar pcluster model propos object similar exhibit coher pattern subset dimens instanc dna microarrai analysi express level gene rise fall synchron respons set environment stimuli magnitud express level close pattern exhibit alik discoveri cluster gene essenti reveal signific connect gene regulatori network commerc applic collabor filter benefit model captur close valu lead indic close purchas brows pattern exhibit custom paper introduc effect algorithm detect cluster perform test real synthet data set show effect mine long sequenti pattern noisi environ applic includ comput biologi studi consum behavior analysi system perform analysi noisi environ observ sequenc accur reflect underli behavior protein sequenc amino acid mutat impact biolog function protein desir occurr observ relat mutat manner support measur number occurr pattern serv purpos paper introduc concept compat matrix mean provid probabilist connect observ underli true metric match propos captur real support pattern expect nois free environ assum addit context address pattern long standard prune techniqu develop market basket problem work effici result algorithm combin statist sampl techniqu border collaps devis discov long pattern minim number scan sequenc databas suffici high confid empir result demonstr robust match model respect nois effici probabilist algorithm tar tempor associ rule evolv numer attribut data mine area increas interest recent year associ rule discoveri problem wide studi unresolv problem research mine pattern evolut numer attribut lack challeng problem signific practic applic busi scienc medicin paper present tempor associ rule model evolv numer attribut metric qualifi tempor associ rule includ familiar measur support strength tradit associ rule mine metric call densiti densiti metric extract rule repres data effect mechan prune search space effici algorithm devis mine tempor associ rule util threshold strength prune search space drastic result rule repres concis manner rule set reduc output size experiment result real synthet data set demonstr effici algorithm meta dpattern reveal hidden period pattern discoveri period pattern time seri data activ research area applic pattern hierarch natur higher level pattern consist repetit lower level pattern presenc nois prevent higher level pattern recogn sens portion data sequenc support high level pattern layout occurr basic symbol exist common represent term raw symbol combin high level pattern express previou model defin raw symbol symbol combin properli recogn exist method paper propos model meta pattern captur high level pattern flexibl model number potenti meta pattern larg substanti difficulti li identifi proper pattern candid apriori properti provid suffici prune power properti compon locat properti identifi conduct candid gener effici comput base mine algorithm develop appli algorithm real synthet sequenc interest pattern discov infomin mine surpris period pattern lessoftheprobabilityofoccurr amoresuitablemeathesensethateveryoccurrencecarriesthesameweightto server theworkloadoneachserverismeasur wardstheassessmentofthesignicanceofapatternregard rank low relativelylow relativelyhigh andhigh thentherear dierentev oneforeach surement inform isintroducedtonaturallyvalueth possiblecombinationofserverst somepreliminari uousandmonotonicallydecreasingfunctionofitsprobabilitydegreeofsurpriseofeachoccurrenceofapatternasacontin ofoccurr thiswouldallowpatternswithvastlydier occurrenceprobabilitiestobehandledseamlessli astheacoccuratamuchlowerfrequencythanotherst pat examinationoftheclusterstatesovertimemightshow thatthestateuctuationcomplieswithsomeperiod cumulateddegreeofsurpriseofallrepetitionsofapattern thealldegreeofsurpriseofthepatternwithinadatasequ conceptofinformationgainisproposedtomeasuretheovertemadministratorsiftheoccurrenceofsuchpattern tern involvingitmaybeofmoreintereststosomesi behavior althoughthehighworkloadonallserversmai theboundedinformationgainpropertyisidentiedtotackl thepredicamentcausedbytheviolationofthedownwardclo earthquak earthquakesoccurveryoftenincalifornia surepropertybytheinformationgainmeasureandinturn demonstratetheeciencyandtheusefulnessofthepropos providesanecientsolutiontothisproblem empiricaltest maybeinterestedinknowingwhetherthereexistsani contradictsthepriorexpect inherentseismicperiodsothatpredictioncanbemad itcanbeclassiedbyitsmagnitudeandtyp scientist model dynam adapt file manag local area network effici mine weight associ rule war paper extend tradit associ rule problem allow weight item transact reflect interest intens item transact turn opportun associ weight paramet item result associ rule call weight associ rule war war improv confid rule mechan effect target market identifi segment custom base potenti degre loyalti volum purchas approach mine war ignor weight find frequent itemset tradit frequent itemset discoveri algorithm introduc weight rule gener shown experiment result approach result shorter averag execut time produc higher qualiti result gener previou method quantit associ rule mine asynchron period pattern time seri data periodici detect time seri data challeng problem great import applic collabor web cach base proxi affin exponenti growth host traffic workload internet collabor web cach recogn effici solut allevi web page server bottleneck reduc traffic cach discoveri locat page cach challeng problem fast grow world wide web environ number particip proxi larg paper propos scheme emploi proxi affin maintain dynam distribut collabor cach infrastructur web page partit cluster proxi refer pattern proxi frequent access page web page cluster form inform group web page belong web page cluster delet ad proxi cach proxi inform group notifi scheme shown greatli reduc number messag overhead individu proxi maintain high cach hit rate final emploi trace driven simul evalu web cach scheme web access trace log verifi cach structur provid signific benefit real workload approach activ spatial data mine base statist inform spatial data mine present challeng due larg size spatial data complex spatial data type special natur spatial access method research area focus effici queri process static data paper introduc activ spatial data mine approach extend current spatial data mine algorithm effici support user defin trigger dynam evolv spatial data exploit local effect updat natur spatial data emploi hierarch structur statist inform level hierarchi decompos user defin trigger set trigger cell hierarchi updat suspendedin hierarchi cumul effect trigger fire shown approach achiev order magnitud improv naiv approach evalu condit databas updat approach produc result delai scheme support increment queri process mine pattern long sequenti data nois dynamo design reliabl massiv scale biggest challeng face amazon largest commerc oper world slightest outag signific financi consequ impact custom trust amazon platform servic web site worldwid implement top infrastructur ten thousand server network compon locat datacent world scale small larg compon fail continu persist state manag face failur drive reliabl scalabl softwar system paper present design implement dynamo highli kei storag system amazon core servic provid experi achiev level avail dynamo sacrific consist failur scenario make extens object version applic assist conflict resolut manner interfac develop sting approach activ spatial data mine spatial data mine present challeng due larg size spatial data complex spatial data type special natur spatial access method research area focus effici queri process static data paper introduc activ spatial data mine approach extend current spatial data mine algorithm effici support user defin trigger dynam evolv spatial data exploit local effect updat natur spatial data emploi hierarch structur statist inform level hierarchi decompos user defin trigger set trigger cell hierarchi updat suspend hierarchi cumul effect trigger fire shown approach achiev order magnitud improv naiv approach evalu condit pk dtree spatial index structur high dimension point data paper present pk tree index structur high dimension point data propos index structur view combin aspect pr quad tree unnecessari node elimin unnecessari node typic result skew point distribut show elimin node perform result index robust skew data distribut index structur formal defin effici updat bound number node height tree prove bound expect height tree mild constraint spatial distribut point choic paramet index structur alloc node page achiev accept storag util low count queri empir evid real data set gener data set rang distribut show pk tree structur deriv consensu algorithm failur detector semin paper chandra toueg show unreli failur detector processor achiev consensu asynchron messag pass system research develop consensu algorithm system base failur detector algorithm develop proven independ paper show consensu algorithm standard model automat convert run result show system model failur detector relat addit permit develop result model transform introduct problem achiev consensu processor distribut system fundament distribut comput consensu achiev presenc failur complet asynchron system messag pass share memori thi dynamo dynam object persist storag map common geoscientif object model heterogen spatial data repositori integr special data manag system geograph inform system gi multimedia system gain import larg varieti data set special repositori user access manipul data set uniform addition desir make special function provid individu repositori user applic homogen interfac ucla data mine laboratori develop geopom geoscientif persist object manag heterogen geoscientif object system homogen interfac heterogen spatial data repositori geopom object orient spatial data model definit user defin spatial object type intern geopom map user defin spatial object type special spatial data repositori emploi storag search spat 