ehuman intellig analysi inform system human cultur dline handwrit recognit parallel machin learn algorithm frequent subgraph mine outerplanar graph recent year increas interest frequent pattern discoveri larg databas graph structur object frequent connect subgraph mine problem tree dataset solv increment polynomi time intract arbitrari graph databas exist approach resort heurist strategi restrict search space identifi practic relev tractabl graph class tree paper class outerplanar graph strict gener tree develop frequent subgraph mine algorithm outerplanar graph show work increment polynomi time practic relev subclass behav outerplanar graph polynomi simpl cycl evalu algorithm empir chemo bioinformat applic list close set strongli access set system applic data mine movement data anonym gener toolkit dbase high dperform data mine larg data mapreduc cluster context dbase cluster imag search result effici discoveri interest pattern base strong closed effici discoveri interest pattern base strong closed tight optimist estim fast subgroup discoveri abstract subgroup discoveri task find subgroup popul exhibit distribut unusu high gener due monoton evalu function standard prune techniqu subgroup discoveri requir optimist estim techniqu optimist estim prune consid extrem simpl case binari target attribut attempt made move suboptim heurist optimist estim paper show optimist estim prune develop sound highli effect prune approach subgroup discoveri base precis definit optim show previou estim tight special case present tight optimist estim popular binari multi class qualiti function present famili increasingli effici approxim optim function show empir experi newli propos optimist estim lead speed order magnitud compar previou approach facilit clinico dgenom knowledg discoveri automat select kdd process analysi clinico genom data pose complex problem machin learn high volum data gener easili select suitabl kdd process problem hand increasingli hard experienc research main idea paper facilit process select repres data set graph base ontolog describ data set attribut appli graph mine method perform similar search measur effect comparison data set graph induc ontolog propos effect propos approach evalu dataset result show ontolog base characterist lead improv character data set facilit clinico dgenom knowledg discoveri automat select kdd process analysi clinico genom data pose complex problem machin learn high volum data gener easili select suitabl kdd process problem hand increasingli hard experienc research main idea paper facilit process select repres data set graph base ontolog describ data set attribut appli graph mine method perform similar search measur effect comparison data set graph induc ontolog propos effect propos approach evalu dataset result show ontolog base characterist lead improv character data set visual analyt method movement data explor analysi larg data set carri pure visual mean requir involv databas technolog computer data process comput analysi method combin technolog method visual facilit synerget work comput human uniqu capabl partner util suggest systemat approach defin method techniqu wai link appropri support work main idea softwar tool prepar visual data human analyst detect type pattern visual displai facilit detect pattern understand type pattern exist data underli phenomenon studi focus data describ movement multipl discret entiti chang posit space preserv integr ident defin type pattern movement data basi abstract model data mathemat function map entiti time spatial posit data transform comput visual techniqu facilit detect type pattern suitabl larg data set possibl larg comput memori constraint visual appli data previous aggreg gener mean databas oper comput techniqu effici close pattern mine strongli access set system semidefinit rank graph problem rank vertic undirect graph prefer relat inconsist prefer data prefer form partial order aim find linear extens conform undirect graph real data inconsist prefer backward edg rank graph problem tackl spectral relax recent shown semidefinit relax offer case solut spectral cluster transduct classif paper investig semidefinit relax rank graph incorpor prefer fix angl metric embed vertic final linear extens obtain random project method experi real world data set show expect improv spectral relax special case rank graph investig problem topolog sort minimum feedback arc set minimum length order problem np hard approxim algorithm minimum length order base semidefinit relax random project result point set effici close pattern mine strongli access set system extend abstract geovisu analyt spatial decis support set research agenda visual analyt tool analysi movement data multi dclass ensembl dbase activ learn frequent subgraph mine outerplanar graph recent year increas interest frequent pattern discoveri larg databas graph structur object frequent connect subgraph mine problem tree dataset solv increment polynomi time intract arbitrari graph databas exist approach resort heurist strategi restrict search space identifi practic relev tractabl graph class tree paper class outerplanar graph strict gener tree develop frequent subgraph mine algorithm outerplanar graph show work increment polynomi time practic relev subclass behav outerplanar graph polynomi simpl cycl evalu algorithm empir chemo bioinformat applic effici dregularis squar regress applic unlabel exampl inexpens easi obtain semisupervis approach utilis exampl reduc predict error paper investig semi supervis squar regress algorithm base learn approach similar semisupervis algorithm base algorithm cubic runtim complex number unlabel exampl handl larger set unlabel exampl devis semi parametr variant scale linearli number unlabel exampl experi show signific error reduct regularis larg runtim improv semi parametr approxim propos distribut procedur appli collect data singl site frequent subgraph mine outerplanar graph recent year increas interest frequent pattern discoveri larg databas graph structur object frequent connect subgraph mine problem tree dataset solv increment polynomi time intract arbitrari graph databas exist approach resort heurist strategi restrict search space identifi practic relev tractabl graph class tree paper class outerplanar graph strict gener tree develop frequent subgraph mine algorithm outerplanar graph show work increment polynomi time practic relev subclass behav outerplanar graph polynomi simpl cycl evalu algorithm empir chemo bioinformat applic bia dfree hypothesi evalu multirel domain effect rule induct label graph bia dfree hypothesi evalu multirel domain machin learn autom categor classif text predefin categori wit boom interest ten year due increas avail document digit form ensu organ research commun domin approach problem base machin learn techniqu gener induct process automat build classifi learn set preclassifi document characterist categori advantag approach knowledg engin approach consist manual definit classifi domain expert good effect consider save term expert labor power straightforward portabl domain survei discuss main approach text categor fall machin learn paradigm discuss detail issu pertain problem document represent classifi construct classifi evalu kernel predict graph mine bia dfree hypothesi evalu multirel domain compar studi method reduc myopia hill dclimb search multirel learn hill climb search commonli search algorithm ilp system permit gener theori short run time drawback greedi search strategi myopia graph kernel hard result effici altern abstract real world data structur research kernel method begun investig kernel kind structur data wide tool model structur data graph interest import challeng investig kernel instanc repres graph specif graph tree string consid paper investig kernel label direct graph gener structur shown comput strictli posit definit graph kernel hard solv graph isomorph problem shown comput product featur space index graph featur count number subgraph isomorph graph np hard hand product altern featur space base walk graph comput polynomi time kernel defin paper cyclic pattern kernel predict graph mine graph kernel hard result effici altern abstract real world data structur research kernel method begun investig kernel kind structur data wide tool model structur data graph interest import challeng investig kernel instanc repres graph specif graph tree string consid paper investig kernel label direct graph gener structur shown comput strictli posit definit graph kernel hard solv graph isomorph problem shown comput product featur space index graph featur count number subgraph isomorph graph np hard hand product altern featur space base walk graph comput polynomi time kernel defin paper learn minesweep multirel learn minesweep person game decept easi plai averag human perform optim plai game requir logic arithmet probabilist reason base spatial relationship board simpli check board state consist np complet problem difficulti hand craft strategi plai game ai research interest automat learn strategi experi paper show integr techniqu gener purpos learn system mio result system capabl induc minesweep plai strategi beat win rate averag human player addit discuss background knowledg present experiment result demonstr gain obtain techniqu show strategi learn game compar evalu featur set evolut strategi multirel boost boost establish success techniqu decreas gener error classif learner base predict ensembl hypothes previou research shown techniqu made work cientli context multirel learn simpl learner activ featur select approach reli simpl static method determin featur select order priori ad featur forward manner paper investig distribut inform present boost usefulli exploit learn reweight featur fact dynam adapt featur set ad relev featur remov longer need preliminari result show inform featur set evolut strategi surprisingli mix ect number featur ultim ensembl result classif accuraci compar evalu approach proposition proposition shown promis approach robustli ectiv handl relat data set knowledg discoveri paper compar date method proposition main group logic orient databaseori techniqu experi learn task ilp benchmark task recent intern data mine competit show group specif advantag logic orient method handl complex background knowledg provid express order model databas orient method cient larger data set obtain accuraci vari combin featur produc group valuabl ventur multirel data mine workshop report multirel data mine workshop report featur select proposition scale boost margin dbase inclusionof featur relat macro doper multirel learn search dspace reduct techniqu refin oper frequent area multirel learn induct logic program ilp order search systemat gener order claus correct theori claus reachabl finit number applic refin oper consid learn system refin oper refin oper determin search space system ecienc reason refin oper comput smallest set claus find correct theori stabil ddriven learn system case studi multirel learn popular deal complex learn exampl proce driven fashion past research shown driven approach learn structur complex gener difficult find multirel learn ilp algorithm hand qualiti learn result driven learn depend order exampl stabil issu receiv attent paper present empir result multirel applic domain show instabl affect perform driven ilp system time examin solut instabl problem present algorithm reli stochast select exampl parallel search show algorithm elimin instabl driven search limit addit effort scalabl constant dmemori sampl algorithm pattern discoveri larg databas data mine task instanc problem find interest util function pattern larg databas recent year signific progress achiev scale algorithm task larg databas sequenti sampl techniqu sampl base greedi algorithm give absolut qualiti guarante scalabl exist approach problem respect data respect size pattern space univers assum entir hypothesi space fit main memori paper describ class algorithm extend hypothesi space fit memori maintain algorithm precis epsilon delta qualiti guarante present constant memori algorithm task prove possess requir properti empir comparison compar variabl memori constant memori sampl find interest pattern databas quickli sequenti sampl discoveri problem subgroup associ rule discoveri natur cast hypothes problem goal hypothes hypothesi space score util function present sampl algorithm solv problem issu small number databas queri guarante precis bound con denc qualiti solut sampl approach treat singl hypothesi select problem assum util averag exampl function case frequent util function show algorithm work util estim bound error provid error bound result worst case sampl bound frequent util prove sampl algorithm popular class util function estim bound error algorithm sequenti sens start return discard hypothes good bad exampl algorithm faster worst case bound lern hidden markov model inform extract activ partial label text discoveri deep wide dorder structur case studi domain mutagen scalabl search sampl smart algorithm activ discoveri increment maxim dinstanc daverag util function applic knowledg discoveri problem data independ sampl bound grossli overestim amount data need individu problem instanc led signific recent interest sequenti algorithm give precis guarante qualiti result determin amount data need base characterist actual problem instanc hand significantli fewer exampl paper present practic sequenti sampl algorithm capabl quickli identifi hypothes work util function estim bound error algorithm predict learn task associ rule find subgroup discoveri experi real world domain show algorithm order magnitud faster sequenti sampl algorithm mine web activ hidden markov model introduct enorm amount inform unstructur semi structur textual document tool inform extract enorm import tool identifi relev inform document convert structur format databas xml document algorithm hand craft set rule research turn learn extract rule hand label document rule base approach fail provid robust inher variabl document structur led recent interest hidden markov model hmm purpos speech recognit comput biochemistri applic hmm markov model algorithm part ofspeech tag hidden markov model inform extract requir train document increment maxim dinstanc daverag util function applic knowledg discoveri problem data independ sampl bound grossli overestim amount data need individu problem instanc led signific recent interest sequenti algorithm give precis guarante qualiti result determin amount data need base characterist actual problem instanc hand significantli fewer exampl paper present practic sequenti sampl algorithm capabl quickli identifi hypothes work util function estim bound error algorithm predict learn task associ rule find subgroup discoveri experi real world domain show algorithm order magnitud faster sequenti sampl algorithm activ hidden markov model inform extract transform dbase learn multirel aggreg relat learn constrain confid drate boost proposit learn boost popular techniqu increas accuraci classif learner firstord learn hand surprisingli attent paid boost due fact simpl form boost lead loss comprehens slow standard ilp learner paper show concern address recent propos techniqu constrain confidencer boost fast weak ilp learner give detail descript algorithm show standard benchmark problem weak learner boost perform compar state art ilp system maintain accept comprehens obtain short run time scalabl search sampl smart algorithm activ discoveri relat instanc dbase learn list term extend dmean cluster dorder represent sequenti sampl algorithm gener class util criteria discoveri problem subgroup associ rule discoveri natur cast hypothesi problem goal hypothes hypothesi space score util function present sampl algorithm solv problem issu small number databas queri guarante precis bound condenc qualiti solut sampl algorithm assum util averag exampl function case frequent util function show algorithm work util estim bound error provid error bound result worst case sampl bound frequent util prove sampl algorithm popular class util function algorithm sequenti sens start return discard hypothes applic learn method hungarian part dof dspeech tag point view comput linguist hungarian difficult languag due complex grammar rich morpholog mean common task part speech tag present challeng learn look hungarian languag fact languag fairli free word order paper present case studi design illustr potenti limit current ilp ilp algorithm hungarian po tag task select popular progol system proposit ilp repres ad experi method aglearn preprocessor base attribut grammar ilp approach phm ribl system compar hungarian version multilingu morphosyntact annot multext east telri corpu consist token experiment result hungarian po tag ind scalabl issu induct logic program measur similar rna structur relat instanc dbase learn step detect rna signal structur silico term comparison dorder similar measur relat distanc dbase cluster work order cluster primarili focus task conceptu cluster form cluster symbol gener represent languag contrast proposit represent experi shown simpl algorithm base exclus distanc measur outperform concept base counterpart paper build recent advanc area rst order distanc metric present rdbc bottom agglom cluster algorithm rst order represent reli distanc inform featur paramet free prune measur select nal cluster cluster tree algorithm empir shown produc good cluster mutagenesi domain subsequ predict task improv previou cluster result approach accuraci dedic predict learner term comparison dorder similar measur data mine serviceteil einsatz von data mine dtechniken zur analys kologisch standort und pflanzendaten algorithm multi drelat discoveri subgroup problem find statist unusu subgroup multi relat databas extend previou work singlerel subgroup discoveri give precis definit multirel subgroup discoveri task propos specif form declar bia base foreign link mean hypothesi space show proposit evalu function adapt multi relat set describ algorithm problem set optimist estim minim support prune optim refin oper sampl ensur effici easili parallel interact configur kikon data mine da aktuel schlagwort extens data mine system success applic data mine techniqu ideal requir system support entir knowledg discoveri process analysi algorithm task hand number success data mine system support entir mine process limit fix select analysi algorithm paper argu favor extens kei featur data mine system discuss requir entail system architectur identifi point exist data mine system fail meet requir describ integr architectur data mine system address problem base concept plug in kepler data mine system built architectur present discuss keyword data mine system architectur extens kepler introduct data mine knowledg discoveri databas kdd aim indukt logikprogrammierung grundlagen und techniken machin learn ecml util problem learn system occur knowledg learn attempt improv system perform degrad perform present methodolog analysi util problem comput model problem solv system isol root util problem detect threshold condit problem aris design strategi elimin present model case base reason control rule learn system compar perform respect swamp util problem analysi suggest case base reason system resist util problem control rule learn system introduct interest asymmetri exist pattern retriev case base reason cbr control rule learn crl system advantag past learn experi crl system retriev rule memori step cbr system ne concept format interact theori revis proper definit minim special theori revis central oper increment learn system special incorrect theori order exclud incorrect infer paper discuss properti requir theori revis oper examin revis minim surpris result seemingli natur criterion requir revis produc maxim gener correct special lead number problem propos altern interpret minim base notion base revis theori contract work formal defin set base revis postul present revis oper mbr meet postul shown produc maxim gener correct revis satisfi postul revis produc mbr minim sens oper implement krt knowledg revis tool mobal panel evalu chang represent concept acquisit model ground concept format research concept format machin learn cognit psycholog featur concept built assum provid elementari vocabulari paper argu unnecessarili limit paradigm examin concept format base evid psycholog machin learn contend principl account origin featur ground model concept format model incorpor direct access world sensor manipul discuss domain process control suitabl framework research model present approach problem develop elementari vocabulari perceptu sensor data die umweltverankerung von begriffsbildungsprozessen automat represent adjust observ discoveri system design goal sloppi model system higher dorder concept tractabl knowledg represent demand ddriven concept format valid rule learn larg high dimension data set interest user domin model data set suppress interest local pattern search interest local pattern implement step learn approach acquir global model focus rest order detect local pattern 