activ infer collect classif label node network import problem grow interest number method exploit local relat inform develop task acquir label node infer time greatli improv accuraci question figur node label acquir challeng previou approach base simpl structur properti present techniqu refer reflect correct learn predict underli classif system make mistak suggest acquisit correct mistak learn algorithm link predict base chanc constraint abstract paper link predict problem partial snapshot network time goal predict addit link form time accuraci current predict method low due extrem class skew larg number potenti link describ learn algorithm base chanc constrain program show exhibit properti need good link predictor preferenti bia posit neg class handl skew data scale larg network experiment result real world domain authorship network biolog network citat network show signific perform improv baselin algorithm conclud briefli describ promis futur direct base work activ learn network data introduc activ learn algorithm classif network data set train instanc connect set link form network label link node correl goal exploit depend accur label node problem aris domain includ social biolog network analysi document classif recent interest method collect classifi node network case label exampl expens network inform show activ learn algorithm advantag network structur algorithm effect exploit link instanc interact local collect aspect classifi improv accuraci learn fewer label exampl experi real world benchmark collect classif domain show achiev extrem accur result small fraction data label graph identif grow tree forest construct folksonomi integr structur metadata social web site user annot content descript metadata tag recent organ content hierarch type structur metadata provid valuabl evid learn commun organ knowledg instanc aggreg person hierarchi common taxonomi folksonomi aid user visual brows social content organ content learn social metadata present challeng spars shallow ambigu noisi inconsist describ approach folksonomi learn base relat cluster exploit structur metadata contain person hierarchi approach cluster similar hierarchi structur tag statist increment weav deeper bushier tree studi folksonomi learn social metadata extract photoshar site flickr demonstr propos approach address challeng compar previou work approach produc larger accur folksonomi addit scale construct folksonomi integr structur metadata social web site user annot content descript metadata tag recent organ content hierarch type structur metadata provid valuabl evid learn commun organ knowledg instanc aggreg person hierarchi common taxonomi folksonomi aid user visual brows social content organ content learn social metadata present challeng spars shallow ambigu noisi inconsist describ approach folksonomi learn base relat cluster exploit structur metadata contain person hierarchi approach cluster similar hierarchi structur tag statist increment weav deeper bushier tree studi folksonomi learn social metadata extract photoshar site flickr demonstr propos approach address challeng compar previou work approach produc larger accur folksonomi addit scale ai theori practic discuss hard challeng opportun ahead indirect dside rel rank robust similar measur gene express data integr structur metadata relat affin propag probabilist approach learn folksonomi structur data read donc function queri evalu probabilist databas probabilist databas hold promis viabl mean larg scale uncertainti manag increasingli need number real world applic domain queri evalu probabilist databas remain comput challeng prior work effici exact queri evalu probabilist databas larg concentr queri centric formul safe plan hierarch queri characterist queri data databas easi construct exampl supposedli hard queri run databas rise tractabl queri evalu problem paper develop effici queri evalu techniqu leverag characterist queri data databas focu tupl independ databas queri evalu problem equival comput margin probabl boolean formula result tupl queri evalu easi boolean formula factor form variabl appear call read suggest naiv approach incorpor previous develop boolean formula factor algorithm queri evalu develop effici factor algorithm work larg subclass queri specif conjunct queri join exploit uniqu structur result tupl boolean formula empir demonstr propos techniqu order magnitud faster gener infer algorithm evalu gener read function special case hierarch queri rival effici prior techniqu specif design handl queri identifi graph noisi incomplet data grow wealth data describ network type includ social network physic network transport commun network biolog network time grow interest analyz network order uncov gener law govern structur evolut pattern predict model develop polici practic fundament challeng deal newli observ data describ network data dubiou qualiti noisi incomplet analysi method appli data clean miss inform infer paper introduc notion graph identif explicitli model infer clean output network noisi input graph output network analysi present illustr explor type infer involv graph identif challeng issu involv combin infer present simpl gener approach combin infer graph identif experiment show util combin approach perform graph identif sensit inter depend infer proceed st acm sigkdd workshop knowledg discoveri uncertain data dynam actor loyalti group affili network dynam actor loyalti group affili network supervis unsupervis method emploi discours relat improv opinion polar classif work investig design choic model discours scheme improv opinion polar classif divers global infer paradigm supervis collect classif framework unsupervis optim framework approach perform substanti baselin approach establish efficaci method underli discours scheme present quantit qualit analys show improv achiev distinguish knowledg social capit social media role context social media commun wikipedia flickr live amp give rise distinct type content foremost relat content discuss chat factual content answer question problem solv user research increasingli interest develop strategi rapidli distinguish type content text base structur strategi extend bodi research show social context social role answer predict content type test framework dataset manual label contribut microsoft live amp find reliabl extract factual relat messag data devolut social affili network work address problem model social network gener explain link group format recent studi social network evolut propos gener model captur statist properti real world network relat node node link format propos model captur coevolut social affili network provid surpris insight group format base observ real world network show user join group reason friend experi show model captur newli observ previous studi network properti work propos gener model captur statist properti complex network propos model facilit control experi studi effect actor behavior network evolut gener realist synthet dataset identifi graph noisi incomplet data grow wealth data describ network type includ social network physic network transport commun network biolog network time grow interest analyz network order uncov gener law govern structur evolut pattern predict model develop polici practic fundament challeng deal newli observ data describ network data dubiou qualiti noisi incomplet analysi method appli data clean miss inform infer paper introduc notion graph identif explicitli model infer clean output network noisi input graph output network analysi present illustr explor type infer involv graph identif challeng issu involv combin infer present simpl gener approach combin infer graph identif experiment show util combin approach perform graph identif sensit inter depend infer futurerank rank scientif articl predict futur pagerank futurerank rank scientif articl predict futur pagerank maximum coverag stream model applic multi dtopic blog dwatch gener graph stream model hypergraph stream model hyperedg arriv onlin comput fly small amount space hyperedg view set element node refer propos model set stream model comput problem maximum coverag set select maxim total weight cover element set stream model comput show algorithm maximumcoverag achiev approxim factor multipl pass allow provid log approxim algorithm set cover multi topic blog watch applic extens blogalert applic handl simultan multipletop request show problem maximumcoverag set cover set stream model util give effici onlin solut problem verifi effect method synthet real weblog data join join illus privaci social network mix public privat user profil order address privaci concern social media websit user hide person profil public work show adversari exploit onlin social network mixtur public privat user profil predict privat attribut user map problem relat classif problem propos practic model friendship group membership inform hidden infer sensit attribut kei idea addit friendship link group carrier signific inform show social media site easili accur recov inform privat profil user knowledg work link base group base classif studi privaci implic social network mix public privat user profil index interact physic design tune model kei task databas administr optim set materi indic respect current workload aid administr challeng task commerci dbmss provid advisor recommend set indic base sampl workload left administr decid recommend indic materi decis requir knowledg indic benefit workload difficult understand depend interact indic advisor provid crucial inform part recommend motiv shortcom propos framework tool administr understand interact recommend set indic formal notion index interact develop algorithm identifi interact relationship exist set indic present experiment result prototyp implement ibm db demonstr effici approach describ databas tune tool util inform index interact tool visual interact base partit index set interact subset tool comput schedul materi indic mainten window maxim benefit case provid strong analyt result show index interact enabl enhanc function summari acm sigkdd workshop knowledg discoveri uncertain data prdb manag exploit rich correl probabilist databas reflect correct misclassif predict approach activ infer prdb manag exploit rich correl probabilist databas probabilist abstract paper present bayesian method construct probabilist network databas focu construct bayesian belief network potenti applic includ comput assist hypothesi test autom scientif discoveri autom construct probabilist expert system extend basic method handl miss data hidden latent variabl show perform probabilist infer averag infer multipl belief network result present preliminari evalu algorithm construct belief network databas case final relat method paper previou work discuss open problem effect label acquisit collect classif friendship ti famili circl link predict leverag social context search social media trust spam report report dbase reput system email filter collect classif network data real world applic produc network data worldwid web hypertext document connect hyperlink social network peopl connect friendship link commun network comput connect commun link biolog network protein interact network recent focu machin learn research extend tradit machin learn classif techniqu classifi node network articl provid introduct area research prefac part seri studi support california governor advisori committe educ excel prefac list figur iv learn structur bayesian network combin abstract hierarchi tree dstructur condit probabl tabl cost dsensit learn condit markov network interact entiti resolut relat data visual analyt tool evalu structur machin learn ten year field induct logic program ilp made steadi progress ilp workshop base balanc develop theori implement applic recent increas emphasi probabilist ilp relat field statist relat learn srl structur predict goal current paper emerg trend chart strateg direct open problem broader area structur machin learn year exploit share correl probabilist databas recent surg work probabilist databas propel larg part huge increas noisi data sourc sensor data experiment data data uncur sourc grow databas manag system effici repres queri data work show data characterist leverag make queri evalu process effici exploit refer share correl uncertainti correl occur repeatedli data share correl occur due reason uncertainti correl gener statist rare vari tupl tupl basi queri evalu procedur introduc correl prior work shown queri evalu problem probabilist databas equival probabilist infer problem appropri construct probabilist graphic model pgm leverag introduc data structur call random variabl elimin graph rv elim graph built pgm obtain queri evalu develop techniqu base bisimul compress rv elim graph exploit presenc share correl pgm compress rv elim graph run infer valid method evalu empir show share correl signific speed up voila effici featur dvalu acquisit classif onlin collect entiti resolut relationship identif social network discoveri recent year inform onlin commun transform wai connect collabor friend colleagu million individu commun onlin dai uniqu opportun observ format evolut role relationship network group organ number challeng aris attempt infer underli social network data ambigu incomplet context depend paper problem collabor network discoveri domain intellig analysi litig support analyst attempt construct valid represent social network specif address challeng relationship identif object identifi relev commun substanti social relationship type propos supervis rank approach problem assess perform manag subordin relationship identif task enron email corpu exploit messag content ranker routin cue analyst relev commun relationship messag traffic indic social relationship dual dview approach interact network visual visual network data tree structur arbitrarili connect graph difficult problem inform visual larg part problem network data user visual attribut specif data item link item connect past approach resolv difficulti focu zoom cluster filter appli method lai node edg approach focu optim network visual singl view limit amount inform shown explor parallel past approach user cross refer subset aspect larg complex network paper propos approach multipl coordin view network illustr approach implement tool call dualnet evalu tool case studi email commun network show multipl coordin view improv navig insight larg network multipl node link type combin collect classif link predict repres tupl attribut uncertainti probabilist databas recent surg work probabilist databas propel larg part huge increas noisi data sourc sensor data experiment data data uncur sourc grow flexibl repres uncertainti data effici queri data build exist probabilist databas work present unifi framework flexibl represent correl tupl attribut level uncertainti import capabl represent abil repres share correl structur data provid motiv exampl illustr share correl structur exist repres share correl structur sophist infer techniqu base lift probabilist infer turn achiev signific speedup comput probabl result user submit queri character rna secondari dstructur featur effect splice dsite predict abstract rna molecul distinguish sequenc composit dimension shape call secondari structur secondari structur pre mrna sequenc strong influenc gene splice previou work show splice site model emploi sequenc featur built featur gener algorithm effect predict splice site gener sequenc featur contain biolog relev featur paper extend featur gener algorithm construct secondari structur featur featur captur nucleotid pair tendenc splice site neighborhood extend splice site model includ pre mrna sequenc structur characterist model significantli outperform sequenc base featur model identifi secondari structur featur captur biolog relev signal splice silenc found signal prefer specif region splice site neighborhood detail prefer geoddup interfac interact entiti resolut geospati data due grow interest geospati data mine analysi data clean integr geospati data import issu geospati entiti resolut process reconcil multipl locat refer real world locat singl data sourc dedupl multipl data sourc integr paper introduc interact tool call geoddup effect combin automat data mine algorithm geospati entiti resolut network visual support user resolut analysi decis illustr geoddup interfac geospati dataset show user effici accur resolv locat entiti final case studi real world geospati dataset demonstr potenti geoddup visual mine multi dmodal social network abstract level social network continu featur rich local global structur properti descript attribut sophist social network analysi support visual mine task number visual tool social network applic develop limit uni modal graph represent tool support wide rang visual option includ interact view support calcul structur graph properti densiti graph deploi tradit statist social network analysi present invenio tool visual mine social invenio integr wide rang interact visual option prefus graph mine algorithm support jung integr expand breadth function core engin tool goal interact explor multi modal multi relat social network invenio support construct view databas oper basic graph mine oper graph identif abstract collect probabilist preserv privaci sensit relationship graph data abstract paper focu problem preserv privaci sensit relationship graph data refer problem infer sensit relationship anonym graph data link reidentif propos privaci preserv strategi vari term amount data remov util amount privaci preserv assum adversari accur predict model link show experiment success link identif strategi vari structur characterist data abstract collect probabilist leverag data structur ontolog integr probabilist interv xml abstract interest xml databas grow year paper studi problem incorpor probabilist inform xml databas propos probabilist interv xml pixml short data model paper data model user express probabilist inform xml markup addit provid altern formal model theoret semant pixml data semant global semant intuit directli amen comput semant local semant amen comput prove result link semant knowledg formal model theoret semant probabilist interv xml provid oper semant comput answer queri correct larg class probabilist instanc featur gener comput splice dsite predict correspond function element increas predict power affili network queri dtime entiti resolut entiti resolut problem reconcil databas refer real world entiti abund publicli databas unresolv entiti motiv problem queri time entiti resolut quick accur resolut answer queri unclean databas queri time collect entiti resolut approach relat refer resolv jointli shown accur independ attribut base resolut line entiti resolut focu develop algorithm collect resolut answer entiti resolut queri queri time purpos formal show collect resolut precis recal individu entiti follow geometr progress neighbor increas distanc consid unfold progress lead natur stage expand resolv queri process strategi strategi extract relat record queri expans oper resolv extract record collect show strategi adapt queri time entiti resolut identifi resolv databas refer help process queri valid approach larg real world public databas show us collect resolut time demonstr adapt strategi queri process show queri answer real time adapt approach preserv gain collect resolut addit experi real dataset synthet gener data empir demonstr valid perform trend predict analysi collect entiti resolut wide rang structur characterist data spliceport interact splice dsite analysi tool visual analysi dynam group membership tempor social network group tool analyz dynam group membership tempor social network time unlik network visual tool show group structur entir network group membership singl actor group user focu analysi pair individu group view addit delet node actor edg relationship time major contribut focu chang group membership time user investig context tempor group membership pair group user flexibl interfac defin redefin group interact support visual represent evolv group membership flexibl user altern view network size user insight group behavior demonstr util tool scientif public network collect entiti resolut relat data databas uncertain imprecis refer real world entiti absenc identifi underli entiti result databas multipl refer entiti lead data redund inaccuraci queri process knowledg extract problem allevi entiti resolut entiti resolut involv discov underli entiti map databas refer entiti tradition entiti resolv pairwis similar attribut refer addit relat inform data specif refer entiti cooccur case collect entiti resolut entiti cooccur refer determin jointli independ improv entiti resolut accuraci propos relat cluster algorithm attribut relat inform determin underli domain entiti give effici implement investig impact relat similar measur entiti resolut qualiti evalu collect entiti resolut algorithm multipl real world databas show improv entiti resolut perform attribut base baselin algorithm relat inform resolv entiti collect addit perform detail experi synthet gener data identifi data characterist favor collect relat resolut pure attribut base algorithm probabilist abstract paper present bayesian method construct probabilist network databas focu construct bayesian belief network potenti applic includ comput assist hypothesi test autom scientif discoveri autom construct probabilist expert system extend basic method handl miss data hidden latent variabl show perform probabilist infer averag infer multipl belief network result present preliminari evalu algorithm construct belief network databas case final relat method paper previou work discuss open problem entiti resolut geospati data integr social capit friendship devent network cost dsensit learn condit markov network queri dtime entiti resolut entiti resolut problem reconcil databas refer real world entiti abund publicli databas unresolv entiti motiv problem queri time entiti resolut quick accur resolut answer queri unclean databas queri time collect entiti resolut approach relat refer resolv jointli shown accur independ attribut base resolut line entiti resolut focu develop algorithm collect resolut answer entiti resolut queri queri time purpos formal show collect resolut precis recal individu entiti follow geometr progress neighbor increas distanc consid unfold progress lead natur stage expand resolv queri process strategi strategi extract relat record queri expans oper resolv extract record collect show strategi adapt queri time entiti resolut identifi resolv databas refer help process queri valid approach larg real world public databas show us collect resolut time demonstr adapt strategi queri process show queri answer real time adapt approach preserv gain collect resolut addit experi real dataset synthet gener data empir demonstr valid perform trend predict analysi collect entiti resolut wide rang structur characterist data grand challeng dprize data mine panel discuss excit motiv grand challeng problem data mine focus bioinformat multimedia mine link mine text mine web mine categori subject descriptor databas applic data mine featur gener algorithm sequenc applic splice dsite predict abstract paper present approach featur select sequenc data identifi gener featur categori give construct algorithm show integr system tightli coupl featur construct featur select integr process refer featur gener systemat search larg space potenti featur demonstr effect approach import compon gene find problem splice site predict show predict model built featur gener algorithm achiev signific improv accuraci exist state art approach latent dirichlet model unsupervis entiti resolut entiti resolut receiv consider attent recent year refer underli entiti goal predict refer correspond entiti show extend latent dirichlet alloc model task propos probabilist model collect entiti resolut relat domain refer connect approach differ recent propos entiti resolut approach gener make pair wise decis captur relat entiti hidden group variabl propos sampl algorithm collect entiti resolut unsupervis take entiti relat account addition assum domain entiti show infer number entiti data demonstr util practic relat entiti resolut approach author resolut real world bibliograph dataset addit present preliminari result character condit relat inform refer resolut organiz email archiv onlin commun provid rich resourc understand social network inform actor dynam role relationship infer commun content traffic structur kei compon analysi onlin commun email resolut refer bodi messag refer resolut reli context messag content messag sender recipi relationship resolv refer investig varieti approach make email traffic network disambigu email refer email traffic network serv proxi infer relationship relationship turn infer candid refer initi find suggest simpl tempor model effect resolv refer class model propos perform maxim exploit long term traffic statist rank candid collect entiti resolut relat data databas uncertain imprecis refer real world entiti absenc identifi underli entiti result databas multipl refer entiti lead data redund inaccuraci queri process knowledg extract problem allevi entiti resolut entiti resolut involv discov underli entiti map databas refer entiti tradition entiti resolv pairwis similar attribut refer addit relat inform data specif refer entiti cooccur case collect entiti resolut entiti cooccur refer determin jointli independ improv entiti resolut accuraci propos relat cluster algorithm attribut relat inform determin underli domain entiti give effici implement investig impact relat similar measur entiti resolut qualiti evalu collect entiti resolut algorithm multipl real world databas show improv entiti resolut perform attribut base baselin algorithm relat inform resolv entiti collect addit perform detail experi synthet gener data identifi data characterist favor collect relat resolut pure attribut base algorithm introduct probabilist graphic model relat data collect entiti resolut relat data databas uncertain imprecis refer real world entiti absenc identifi underli entiti result databas multipl refer entiti lead data redund inaccuraci queri process knowledg extract problem allevi entiti resolut entiti resolut involv discov underli entiti map databas refer entiti tradition entiti resolv pairwis similar attribut refer addit relat inform data specif refer entiti cooccur case collect entiti resolut entiti cooccur refer determin jointli independ improv entiti resolut accuraci propos relat cluster algorithm attribut relat inform determin underli domain entiti give effici implement investig impact relat similar measur entiti resolut qualiti evalu collect entiti resolut algorithm multipl real world databas show improv entiti resolut perform attribut base baselin algorithm relat inform resolv entiti collect addit perform detail experi synthet gener data identifi data characterist favor collect relat resolut pure attribut base algorithm prl probabilist relat languag grand challeng data mine kdd panel report bayesian network learn abstract hierarchi context dspecif independ ddupe interact tool entiti resolut social network visual analyz social network challeng problem receiv grow attent import step analysi begin ensur data accur common data qualiti problem data inadvert distinct refer underli entiti process reconcil refer call entityresolut dupe interact tool combin data mine algorithm entiti resolut task specif network visual user cope complex clean larg network focus small subnetwork potenti duplic pair subnetwork highlight relationship social network make common relationship easi visual identifi dupe user resolv ambigu merg node mark distinct entiti resolut process iter pair node resolv addit duplic reveal resolut decis chain give exampl user flexibl appli sequenc action produc high qualiti entiti resolut result illustr evalu benefit dupe bibliograph collect dataset clean contain duplic fact duplic rapidli identifi dupe uniqu combin entiti resolut algorithm task specif visual interfac prune social network structur properti descript attribut scale issu understand make sens larg social network investig method prune social network determin relev relationship measur import term predict accuraci set target attribut social network goal creat prune network model inform affili relationship present method prune network base structur properti descript attribut demonstr network nasdaq nyse busi bibliograph network tutori statist relat learn abstract collect probabilist execut summari probabilist introduct special issu link mine link mine survei unsupervis sens disambigu bilingu probabilist model describ probabilist model unsupervis word sens disambigu parallel corpora model call sens model build work diab resnik parallel text sens inventori target languag recast approach probabilist framework model call concept model hierarch model concept latent variabl relat languag specif sens label show model improv perform word sens disambigu task previou unsupervis approach concept model show largest improv learn concept model product learn sens inventori parallel languag iter record linkag clean integr structur web site automat segment tabl web site dynam gener html page displai result user queri present inform form list tabl current tool applic programmat extract inform reli heavili user input form label extract record sheer size rate growth web make solut reli primarili user input infeas long term fortun web site explicit implicit structur layout content exploit purpos inform extract paper describ approach automat extract segment record web tabl automat method requir user input reli sole layout content web sourc approach reli common structur web site present inform list tabl link entri lead detail page addit inform item describ algorithm redund content tabl detail page aid inform extract algorithm encod addit inform provid detail page constraint find segment solv constraint satisfact problem algorithm probabilist infer find record segment show approach exploit web site structur gener domain independ manner demonstr effect algorithm set twelv web site understand tuberculosi epidemiolog structur statist model workshop report mine semant web kdd proceed ninth acm sigkdd intern confer knowledg discoveri data mine problem identifi approxim duplic record databas essenti step data clean data integr process exist approach reli gener manual tune distanc metric estim similar potenti duplic paper present framework improv duplic detect trainabl measur textual similar propos emploi learnabl text distanc function databas field show measur capabl adapt specif notion similar field domain present learnabl text similar measur suitabl task extend variant learnabl string edit distanc vector space base measur emploi support vector machin svm train experiment result rang dataset show framework improv duplic detect accuraci tradit techniqu pxml probabilist semistructur data model algebra ehung getoor recent prolifer work semistructur data model work date support uncertainti model paper propos model probabilist semistructur data psd advantag approach support flexibl represent specif wide class distribut semistructur instanc provid semant model show semant probabilist coher develop extens relat algebra handl probabilist semistructur data describ effici algorithm answer queri algebra final present experiment result show effici algorithm link dbase classif surg interest learn mix label unlabel data gener approach includ semi supervis learn tranduct infer paper uniqu wai unlabel data improv perform link base classif classif object make object descript link object learn structur statist model link mine data mine challeng learn probabilist model link structur real world data heterogen richli interconnect exampl includ web hypertext bibliometr data social network contrast statist learn method work flat data represent forc convert data form lose link structur recent introduc framework probabilist relat model prm embrac object relat natur structur data captur probabilist interact attribut relat entiti paper extend framework model interact attribut link structur advantag approach unifi generar model content relat structur propos mechan repres probabilist distribut link structur refer uncertainti exist uncertainti describ condit model present learn algorithm present experiment result show learn model predict link structur observ link structur provid predict attribut model learn probabilist model relat structur real world data store relat form contrast statist learn method work flat data represent forc convert data form lose relat structur recent introduc framework probabilist relat model prm repres probabilist model multipl entiti util relat paper propos probabilist model attribut relat model relat structur propos mechan model structur uncertainti refer uncertainti exist uncertainti describ condit model present learn algorithm present experiment result show learn model predict relat structur observ relat structur provid predict attribut model select estim probabilist model estim result size complex queri involv select multipl attribut join relat difficult fundament task databas queri process aris cost base queri optim queri profil approxim queri answer paper show probabilist graphic model effect task accur compact approxim joint frequenc distribut multipl attribut multipl relat probabilist relat model prm recent develop extend graphic statist model bayesian network relat domain repres statist depend attribut tabl attribut foreign kei join provid effici algorithm construct prm databas show prm comput select estim broad class queri major contribut work unifi framework estim queri involv select foreign kei join oper approach limit answer small set predetermin queri singl model effect estim size wide collect potenti queri multipl tabl present result approach real world databas singl tabl multi attribut queri gener class select join queri approach produc accur estim standard approach select estim compar space time aaai workshop report featur hierarchi bayesian network learn recent year research statist uai commun develop impress bodi theori algorithm machineri learn bayesian network data learn bayesian network pattern discoveri predict diagnosi densiti estim task earli pioneer work area includ algorithm emerg current popular approach simpl greedi hill climb algorithm search space candid structur guid network score function bayesian minimum descript length mdl base search begin initi candid network typic empti network edg consid make small local learn probabilist relat model larg portion real world data store commerci relat databas system contrast statist learn method work amp quot flat amp quot data represent appli method forc convert data flat form lose relat structur present databas paper build recent work probabilist relat model prm describ learn databas prm properti object depend probabilist properti object properti relat object prm significantli express standard model bayesian network show extend statist method learn bayesian network learn model describ paramet estim structur learn automat induct depend structur model show learn procedur exploit standard databas retriev techniqu effici learn larg dataset present experiment result real synthet relat databas learn probabilist relat model larg portion real world data store commerci relat databas system contrast statist learn method work amp quot flat amp quot data represent appli method forc convert data flat form lose relat structur present databas paper build recent work probabilist relat model prm describ learn databas prm properti object depend probabilist properti object properti relat object prm significantli express standard model bayesian network show extend statist method learn bayesian network learn model describ paramet estim structur learn automat induct depend structur model show learn procedur exploit standard databas retriev techniqu effici learn larg dataset present experiment result real synthet relat databas util elicit classif problem investig applic classif techniqu util elicit decis problem set paramet gener elicit probabl util prior condit probabl model chang user user util model elicit util model separ user elicit effect redund constraint onlin schedul heurist mean improv constraint solver perform research wide work problem independ heurist variabl order focus offlin problem shot constraint satisfact paper present onlin schedul problem develop real time schedul algorithm gener heurist schedul focu domain specif redund constraint effect approxim optim offlin solut present taxonomi redund domain constraint examin impact effect schedul describ techniqu gener redund constraint appli larg class job shop schedul problem scope abstract criteria local plan ion criteria local plan ami lanski lise getoor recom technolog nasa am research center artifici intellig research branch ms moffett field ca lanski ptolemi arc nasa gov getoor ptolemi arc nasa gov abstract local gener purpos represent techniqu partit problem subproblem local problem solver search smaller search space subproblem unlik method partit local subproblem overlap multipl search space involv construct share piec plan paper focu criteria form local scope abstract describ method automat gener local provid empir result contrast offic build construct domain introduct year research focus abstract 