commun motif tool character social commun comput label dconstraint reachabl graph databas dens pattern mine graph stream massiv web commun networkappl creat data repres massiv sequenti stream edg convers telecommun network messag social network repres massiv stream edg stream typic larg larg amount underli activ network import applic domain determin frequent occur dens structur underli graph stream gener determin frequent dens pattern underli interact introduc model dens pattern mine propos probabilist algorithm determin structur pattern effect effici purpos probabilist approach creat summar graph stream pattern mine show summar approach lead effect effici result stream pattern mine number real synthet data set effici skylin comput metric space set queri point gener metric space metric space skylin mss queri ask closest point queri point databas point point databas equal distanc queri point denot closest point queri point problem direct gener recent propos spatial skylin queri problem point locat dimension euclidean space close relat nearest neighbor nn queri rang queri common skylin queri problem paper develop algorithm aggress prune skylin point search space contribut optim techniqu reduc number distanc comput domin test experiment evalu shown effect effici approach estim number frequent itemset larg databas estim number frequent itemset minim support larg dataset great interest theoret practic perspect find number frequent itemset number maxim frequent itemset complet studi provid theoret investig sampl estim discov prove fundament surpris properti sampl estim propos algorithm estim number frequent itemset sampl detail experiment result shown accuraci effici propos approach tree dbase framework differ summar sparsif approach tempor graphic model decomposit multipl inform sourc cooper learn applic face problem learn object dataset inform auxiliari sourc benefici integr object dataset learn paper propos omni view learn approach enabl learn multipl data collect theme organ heterogen data sourc unifi tabl global data view achiev omni view learn goal object dataset auxiliari dataset share instanc level depend structur propos relat mean cluster instanc auxiliari dataset cluster build featur captur correl object auxiliari dataset experiment result demonstr omni view learn build model outperform learn object dataset comparison train algorithm assert omni view learn altern effect semi supervis learn migrat motif spatial tempor pattern mine approach financi market recent studi promin financ research fama french introduc framework studi risk return migrat stock size portfolio space financi event attempt disentangl relationship migrat behavior stock return time work deriv result market segment individu compani year move challeng financi data mine captur categor migrat individu compani behavior affect return propos data mine approach studi multi year movement individu compani specif address question discov frequent migrat pattern stock market present trajectori mine algorithm discov migrat motif financi market featur algorithm handl approxim pattern match graph theoret method maxim cliqu identif incorpor tempor spatial constraint perform detail studi nasdaq nyse amex stock market year span successfulli find migrat motif confirm exist financ theori motif lead financi model cartesian contour concis represent collect frequent set cartesian contour concis represent collect frequent set identifi dynam network modul tempor spatial constraint rapid accumul system level biolog data understand dynam natur cellular activ remain difficult task reason biolog data static correspond snapshot cellular activ studi explicitli attempt detangl tempor complex biolog network compil time seri gene express profil data defin dynam network modul set protein satisfi condit form connect compon protein protein interact ppi network express profil form structur tempor domain develop effici mine algorithm discov dynam modul tempor network yeast model system demonstr major identifi dynam modul function homogen addition provid insight sequenti order molecular event cellular system final note applic algorithm limit studi ppi network gener applic combin type network time seri data dhop high dcompress index scheme reachabl queri data discret unif data discret defin process convert continu data attribut valu finit set interv minim loss inform paper prove discret method base inform theoret complex method base statist measur data depend asymptot equival defin notion gener entropi prove discret method base mdlp gini index aic bic pearson statist deriv gener entropi function design dynam program algorithm guarante discret base gener entropi notion conduct extens perform evalu method publicli data set result show method deliv averag classif error previous discret method cost dbase queri optim complex pattern mine multipl databas complex data mine queri queri optim issu aris similar tradit databas queri work appli cost base queri optim kei techniqu optim tradit databas queri complex mine queri work develop cost base queri optim framework import collect data mine queri frequent pattern mine multipl databas specif make contribut present rich class queri mine frequent itemset multipl dataset support sql base mechan present approach enumer queri plan mine queri develop dynam program approach branch bound approach base enumer algorithm find optim queri plan mine cost introduc model estim cost individu mine oper evalu queri optim techniqu real synthet dataset show signific perform improv topic model approach integr random walk framework academ search paper propos unifi topic model approach integr random walk framework academ search specif present topic model simultan model paper author public venu combin propos topic model random walk framework experiment result show propos approach academ search significantli outperform baselin method bm languag model exist topic model includ plsi lda model overlap matrix pattern visual hypergraph approach work studi visual data mine problem set discov overlap submatric interest order row column data matrix displai submatric relationship find problem convert hypergraph order problem gener tradit minim linear arrang graph order problem andthenweareabletoprovethenp hardnessofthisproblem propos iter algorithm util exist graph order algorithm solv optim visual problem algorithm converg local minimum detail experiment evalu set publicli transact datasetsdemonstratestheeffectivenessandefficiencyofth propos algorithm effect effici itemset pattern summar regress dbase approach paper propos set regress base approach effect effici summar frequent itemset pattern specif show problem minim restor error set itemset base probabilist model correspond linear regress problem show condit transform linear regress problem linear regress problem propos method regress tree regress partit entir collect frequent itemset order minim restor error regress approach emploi mean type cluster method guarante total restor error achiev local minimum treeregress approach emploi decis tree type top partitionprocess inaddit wediscussalternativestoestim frequenc collect itemset cover repres itemset experiment evalu realandsyntheticdatasetsdemonstratesthatourapproachessignificantli improv summar perform term bothaccuraci restorationerror andcomput cost succinct summar transact databas overlap hyperrectangl scheme transact data ubiquit method includ frequent itemset mine cluster propos analyz transact databas work propos research problem succinctli summar transact databas solv problem requir link high level structur databas potenti huge number frequent itemset formul problem set cover problem overlap hyperrectangl prove problem variat np hard develop approxim algorithm hy achiev ln approxim ratio polynomi time propos prune strategi significantli speed process algorithm addition propos effici algorithm summar set hyperrectangl allow fals posit condit detail studi real synthet dataset show effect effici approach summar transact databas categori andsubjectdescriptor effici answer reachabl queri larg direct graph effici process queri larg graph import research topic larg driven emerg real world applic divers xml databas gi web mine social network analysi ontolog bioinformat graph reachabl attract lot research attent reachabl queri common graph databas serv fundament oper graph queri main idea answer reachabl queri graph build indic base reachabl label essenti vertex graph assign label reachabl vertic determin label approach propos build reachabl label interv label tree cover hop label due larg number vertic real world graph graph easili million vertic comput cost index size label exist method prove expens practic paper introduc graph structur refer pathtre label larg graph path tree cover span subgraph tree shape demonstr analyt empir effect approach queri plan search inter ddepend deep dweb databas abstract increasingli data sourc onlin databas hidden queri form form refer deep web itisdesir system provid high level simpl interfac user queri data sourc autom data retriev deep web system address challeng case singl databas provid desir data multipl databas queri user queri due depend present deep web databas databas queri databas time network hardwar problem queri plan capabl deal unavail databas gener altern plan optim feasibl paper consid queri plan context deep web integr system develop dynam queri planner gener effici queri order base databas depend queri planner select top queri plan develop cost model suitabl queri plan deep web mine implement evalu made context bioinformat system snpminer compar algorithm naiv algorithm optim algorithm show queri algorithm outperform naiv algorithm obtain similar result optim algorithm experi show scalabl system respect number data sourc involv number queri term queri plan search inter ddepend deep dweb databas abstract increasingli data sourc onlin databas hidden queri form form refer deep web itisdesir system provid high level simpl interfac user queri data sourc autom data retriev deep web system address challeng case singl databas provid desir data multipl databas queri user queri due depend present deep web databas databas queri databas time network hardwar problem queri plan capabl deal unavail databas gener altern plan optim feasibl paper consid queri plan context deep web integr system develop dynam queri planner gener effici queri order base databas depend queri planner select top queri plan develop cost model suitabl queri plan deep web mine implement evalu made context bioinformat system snpminer compar algorithm naiv algorithm optim algorithm show queri algorithm outperform naiv algorithm obtain similar result optim algorithm experi show scalabl system respect number data sourc involv number queri term middlewar data mine applic cluster grid paper overview middlewar system develop year address challeng involv develop parallel distribut implement data mine algorithm freerid framework rapid implement data mine engin focus data mine cluster environ freerid base observ parallel version data mine techniqu share similar structur parallel divid data instanc record transact node comput node involv read data instanc arbitrari order process data instanc perform local reduct reduct involv commut associ oper mean result independ order data instanc process local reduct node global reduct perform similar structur exploit middlewar system execut data mine task effici parallel start high level specif techniqu enabl process data set store remot data repositori extend freerid middlewar freerid framework rapid implement data mine engin grid freerid support high level interfac develop data mine scientif data process applic involv data store remot repositori ad function freerid aim abstract detail remot data retriev movement cach applic develop mmi mine multipl inform sourc workshop report graph topolog structur mine scientif articl snpminer domain dspecif deep web mine tool data discret unif data discret defin process convert continu data attribut valu finit set interv minim loss inform paper prove discret method base inform theoret complex method base statist measur data depend asymptot equival defin notion gener entropi prove discret method base mdlp gini index aic bic pearson statist deriv gener entropi function design dynam program algorithm guarante discret base gener entropi notion conduct extens perform evalu method publicli data set result show method deliv averag classif error previous discret method trend motif graph mine approach analysi dynam complex network assign schema label ontolog huerist exploratori tool follow adup studi microarrai experi decomposit dbase probabilist framework estim select xml twig queri abstract paper present approach estim select xml twig queri techniqu answer approxim queri determin optim queri plan complex queri base estim approach reli summari structur occurr statist small twig reli probabilist approach decompos larger twig queri smaller show estim select larger queri conjunct summari inform present evalu strategi decomposit compar work state art select estim approach synthet real dataset experiment result show propos approach effect estim select xml twig queri systemat approach optim complex mine task multipl databas iter process order support process long term goal data mine research build knowledg discoveri data mine system kddm line research provid databas support mine oper freerid dg support applic mine remot sampl dbase estim olap queri import sampl approxim queri process databas environ differ tradit applic sampl databas feasibl collect accur summari statist data addit sampl paper describ set sampl base estim approxim queri process make simpl summari statist greatli increas accuraci sampl base estim estim give tight probabilist guarante estim accuraci suitabl low high dimension data work categor numer attribut inform estim easili gather singl pass make suitabl stream environ freerid dg support applic mine remot fast exact dof dcore distribut dmean cluster cluster wide studi topic data mine mean cluster popular cluster algorithm mean requir pass entir dataset make expens larg disk resid dataset view lot work approxim version mean requir small number pass entir dataset paper present algorithm call fast exact mean cluster fekm typic requir small number pass entir dataset provabl produc cluster center report origin mean algorithm algorithm sampl creat initi cluster center take pass entir dataset adjust cluster center provid theoret analysi show cluster center report comput origin mean algorithm experiment result number real synthet dataset show speedup factor compar mean paper describ evalu distribut version fekm refer dfekm algorithm suitabl analyz data distribut loos coupl machin unlik previou work area dfekm provabl produc result origin mean algorithm experiment result show dfekm option exact cluster distribut data load data run sequenti mean run parallel mean loos coupl configur tightli coupl environ dfekm outperform parallel mean signific load imbal data mine techniqu learn layout flat dfile biolog dataset learn layout biolog dataset semi dautomat algorithm dcore frequent itemset mine stream data extens studi decad paper take approach problem make major contribut present pass algorithm frequent itemset mine determinist bound accuraci requir core summari structur pass algorithm produc fals neg easili extend pass accur algorithm pass algorithm memori effici mine dataset larg number distinct item low support level parallel defect detect categor applic paper present case studi creat parallel scalabl implement scientif data analysi applic focu defect detect categor applic analyz dataset produc molecular dynam md simul parallel applic goal want achiev high parallel effici want creat implement scale disk resid dataset want creat easi maintain modifi implement high level interfac number techniqu organ input data achiev load balanc effici parallel step updat match defect catalog meet goal system call freerid framework rapid implement datamin engin origin develop parallel data mine algorithm carri detail evalu implement main observ experi implement achiev high parallel effici execut time remain proport amount comput dataset disk resid scheme load balanc method parallel updat match defect catalog crucial parallel effici defect categor phase simultan optim complex mine task knowledg cach discov frequent topolog structur graph dataset share memori parallel data mine algorithm techniqu recent technolog advanc share memori parallel machin scalabl oer larg main memori high bu bandwidth emerg good platform data wareh data mine paper focu share memori parallel data mine algorithm commun memori optim parallel data cube construct data cube construct commonli oper data warehous volum data store analyz data warehous amount comput involv data cube construct natur parallel machin oper methodolog detail perform model reduct comput smp machin paper revisit problem perform predict smp machin motiv select parallel strategi random write reduct reduct frequent aris data mine algorithm previou work develop number techniqu parallel class reduct previou work shown techniqu full replic optim full lock cachesensit outperform depend problem dataset machin paramet import question predict perform techniqu problem dataset machin paper address question develop analyt perform model captur level cach coher cach miss tlb miss lock overhead content memori analyt model combin result micro benchmark predict perform real machin valid model smp machin result show model effect captur impact memori hierarchi twolevel cach tlb factor limit parallel content lock memori content coher cach miss differ predict measur perform case model accur predict rel perform parallel techniqu elsevi right reserv tile scale parallel data cube construct parallel data cube construct algorithm present cgmolap server fulli function parallel olap system build data cube rate terabyt hour cgmolap incorpor varieti approach parallel comput full cube partial cube iceberg cube parallel cube index scheme cgmolap system consist applic interfac parallel queri engin parallel cube materi engin meta data cost model repositori share server compon provid uniform manag memori commun disk resourc commun memori optim parallel data cube construct data cube construct commonli oper data warehous volum data store analyz data warehous amount comput involv data cube construct natur parallel machin oper impact data distribut quantil order sequenc data valu element rank theta total number valu accur estim quantil requir solut practic applic paper present algorithm estim quantil valu disk resid data algorithm characterist requir pass data determinist produc good lower upper bound true valu quantil requir priori knowledg distribut data set scalabl parallel formul extra time memori comput addit quantil constant quantil present experiment result ibm sp experiment result show algorithm robust depend distribut data set larg part work khale alsabti vineet sing compil framework distribut memori parallel data mine algorithm effici decis tree construct stream data decis tree construct studi problem data mine recent interest mine stream data domingo hulten present pass algorithm decis tree construct work hoeffd inequ achiev probabilist bound accuraci tree construct paper revisit problem make contribut present numer interv prune nip approach effici process numer attribut result show averag reduct execut time exploit properti gain function entropi gini reduc sampl size requir obtain bound accuraci experiment result show reduct number data instanc requir commun memori effici parallel decis tree construct decis tree construct import data mine problem paper revisit problem goal develop effici parallel algorithm decis tree construct parallel algorithm major mine task implement data cube construct cluster middlewar algorithm implement data cube construct cluster middlewar algorithm design evalu high dlevel interfac data mine share memori parallel decis tree construct gener data mine middlewar design evalu high dlevel interfac data mine compil runtim support share memori parallel data mine algorithm data mine techniqu focu find pattern model larg dataset volum data analyz amount comput involv rapid interact analysi data mine applic requir parallel machin develop compil runtim support develop scalabl implement data mine algorithm work encompass share memori parallel distribut memori parallel optim process disk resid dataset share memori parael data mine algorithm techniqu perform predict random write reduct case studi model share memori program paper revisit problem perform predict share memori parallel machin motiv select parallel strategi random write reduct reduct frequent aris data mine algorithm compil data intens applic spatial coordin process analyz larg volum data plai increasingli import role domain scientif research develop compil process data intens applic written dialect java compil effici execut cluster workstat distribut memori machin effici associ mine implement cluster smp compil middlewar support scalabl data mine compil data intens applic spatial coordin process analyz larg volum data plai increasingli import role domain scientif research develop compil process data intens applic written dialect java compil effici execut cluster workstat distribut memori machin high level program methodolog data intens comput 