exploit local node cach top dk queri wireless sensor network revers dnearest neighbor monitor mobil object techniqu effici search spatial subspac similar search idea rank top dk retriev boost spatial prune optim prune mbr archaeo dinformat scientif data manag archaeobiolog share dneighbor distanc defeat curs dimension share dneighbor distanc defeat curs dimension subspac similar search effici dnn queri arbitrari subspac quest workshop report st intern workshop queri mine uncertain spatio dtempor data seattl metric space data mine applic cluster acm special interest group spatial inform sigspati acm sigspati address issu relat acquisit manag process spatial relat inform focu algorithm geometr visual consider scope includ limit geograph inform system current acm sigspati offic ossobook databas knowledgemanag techniqu archaeozoolog period pattern analysi time seri databas abstract similar search time seri data divers domain promin work focus similar search complet time seri subsequ time seri time seri temperatur measur consist period pattern pattern repeatedli occur defin period time behavior temperatur dai commonli correl dai analysi pattern consecut pattern valuabl applic domain financ medicin meteorolog ecolog paper present framework similar search time seri databas specif period pattern effici threshold base similar search method appli invari small distort time experi real world data show similar measur meaning establish measur applic techniqu effici search spatial revers dnearest neighbor search dynam gener metric databas paper propos origin solut gener revers nearest neighbor rknn search problem compar limit exist method rknn search approach work top hierarch organ tree index structur applic type data long metric distanc function defin data object exemplarili show approach work top preval index structur euclidean metric data tree tree solut applic arbitrari valu appli dynam environ updat databas frequent occur gener solut rknn problem solut outperform exist method term queri execut time exploit strategi prune fals drop identifi true hit constrain revers nearest neighbor search mobil object increment revers nearest neighbor rank abstract paper formal concept increment revers nearest neighbor rank suggest origin solut problem propos effici approach report result increment restart search scratch approach appli multidimension featur databas hierarch organ tree index structur solut assum preprocess step make applic dynam environ updat databas frequent occur experi show approach report rank result page access exist approach design tradit revers nearest neighbor search appli rank problem increment revers nearest neighbor rank abstract paper formal concept increment revers nearest neighbor rank suggest origin solut problem propos effici approach report result increment restart search scratch approach appli multidimension featur databas hierarch organ tree index structur solut assum preprocess step make applic dynam environ updat databas frequent occur experi show approach report rank result page access exist approach design tradit revers nearest neighbor search appli rank problem outlier detect axi dparallel subspac high dimension data abstract propos origin outlier detect schema detect outlier vari subspac high dimension featur space object data set explor axi parallel subspac span neighbor determin object deviat neighbor subspac experi show subspac outlier detect superior exist fulldimension approach scale high dimension databas probabilist similar search uncertain time seri abstract probabilist similar queri uncertain data assign uncertain databas object probabl indic likelihood meet queri predic paper formal notion uncertain time seri introduc import type probabilist rang queri uncertain time seri propos origin approxim represent uncertain time seri effici support queri type upper lower bound euclidean distanc revers dnearest neighbor search base aggreg point access method abstract propos origin solut gener revers nearest neighbor rknn search problem euclidean space compar limit exist method rknn search approach work top multi resolut aggreg mra version index structur multi dimension featur space leaf node addition aggreg inform sum leaf entri index node solut outperform state art rknn algorithm term queri execut time exploit advanc strategi prune index entri increment revers nearest neighbor rank vector space abstract paper formal concept increment revers nearest neighbor rank suggest origin solut problem propos effici approach report result increment restart search scratch approach appli multidimension featur databas hierarch organ tree index structur solut assum preprocess step make applic dynam environ updat databas frequent occur experi show approach report rank result page access exist approach design tradit revers nearest neighbor search appli rank problem multi dstep queri process increas number modern databas applic effici support similar search import task complex object imag molecul mechan part complex similar model increas algorithm directli base index work simpl medium dimension similar distanc function meet effici requir complex high dimension adapt distanc function multi step queri process strategi recommend case investig substanti number candid produc filter step evalu refin step fundament effici paramet reveal strong perform shortcom state art algorithm nearest neighbor search korn al present multi step algorithm guarante produc minim subspac cluster techniqu document cluster fundament task text mine effici organ navig summar retriev document achiev cluster document present difficult challeng due sparsiti high dimension text data complex semant natur languag subspac cluster extens tradit cluster design captur local featur relev group document respect featur word matter paper present subspac cluster techniqu base local adapt cluster lac algorithm improv subspac cluster document identif keyword achiev lac kernel method semant distanc deploi basic idea defin local kernel cluster semant distanc pair word comput deriv cluster local term weight propos approach call semant lac evalu benchmark dataset experi show semant lac capabl improv cluster qualiti subspac project cluster experiment evalu analysi approxim cluster time seri compact model dbase descript abstract cluster time seri limit fact length time seri significantli neg influenc runtim hand approxim cluster appli exist compress represent time seri obtain dimension reduct suffer low accuraci propos method compress time seri base mathemat model explor depend time seri time seri repres combin set specif refer time seri cost represent depend number refer time seri length time seri show small number refer time seri yield accur represent reduc storag cost runtim cluster algorithm significantli experi illustr represent produc approxim cluster high accuraci consider reduc runtim analysi time seri compact model dbase descript abstract recent propos method compress time seri base mathemat model explor depend time seri represent model time seri combin set specif refer time seri cost represent depend number refer time seri length time seri demonstr present java toolkit perform data mine task base time seri represent framework user explor properti approach comparison state art compress method result visual present concis user easili identifi import set model base time seri represent continu proxim monitor road network paper scenario set mobil object continu track posit road network commun central server server posit updat move object detect event object reach exce proxim distanc server perman awar pair object distanc rang commun cost object server quickli bottleneck posit updat server track time slot propos updat strategi order reduc commun overhead defin special region object region defin posit updat server requir long object leav region present effici algorithm updat region detect proxim separ object leav region empir evalu strategi term commun overhead number requir posit updat effici queri process larg traffic network abstract present origin graph embed speedup distanc rang nearest neighbor queri static dynam object locat weight graph method comput lower upper bound filter distanc approxim true shortest path distanc significantli tradit filter addit discuss comput exact shortest path distanc refin step boost embed graph dtime threshold dbase data mine time seri abstract mine time seri data import approach analysi applic area divers biologi environment research medicin stock chart analysi data mine task kind data depend distanc function time seri huge number function develop decad introduct threshold base distanc function present concept time seri similar function appli data mine techniqu wide spectrum time seri data demonstr present java toolkit time perform data mine task complet rang threshold valu interact result visual present concis user easili identifi import threshold valu combin domain specif knowledg pivot valu yield insight mean underli data mine techniqu analysi base similar search multimedia time seri data amplitud dlevel featur robust cluster arbitrarili orient subspac paper propos effici effect method find arbitrarili orient subspac cluster map data space paramet space defin set arbitrarili orient subspac object cluster algorithm base principl find subspac accommod databas object contrast exist approach method find subspac cluster dimension spars intersect cluster noisi environ broad experiment evalu demonstr robust effici effect method hierarch graph embed effici queri process larg traffic network gener framework increas robust pca dbase correl cluster algorithm abstract correl cluster algorithm reli princip compon analysi pca correl analysi tool correl cluster learn appli pca set sampl point pca sensit outlier small fraction point correspond correct correl cluster algorithm misl fail detect correct result paper evalu influenc outlier pca propos gener framework increas robust pca order determin correct correl cluster show framework appli pca base correl cluster algorithm experiment evalu show benefit framework synthet real world data set gener framework increas robust pca dbase correl cluster algorithm abstract correl cluster algorithm reli princip compon analysi pca correl analysi tool correl cluster learn appli pca set sampl point pca sensit outlier small fraction point correspond correct correl cluster algorithm misl fail detect correct result paper evalu influenc outlier pca propos gener framework increas robust pca order determin correct correl cluster show framework appli pca base correl cluster algorithm experiment evalu show benefit framework synthet real world data set detect cluster moder dto dhigh dimension data subspac cluster prolif research area data mine subspac cluster relat problem induc vast amount propos solut public compar proposit competitor call na ve ad hoc solut fail clarifi exact problem definit consequ solut compar experiment remain unclear solut tackl problem agre tacit assumpt assumpt influenc outcom algorithm tutori clarifi problem definit relat subspac cluster gener ii specif difficulti encount field research iii vari assumpt heurist intuit form basi approach iv promin solut essenti tackl problem global correl cluster base hough transform effici revers dnearest neighbor estim abstract revers nearest neighbor rknn problem find object data set nearest neighbor includ queri object receiv increas attent recent industri scientif applic call solut rknn problem arbitrari metric space data object euclidean metric distanc function object similar applic solut gener problem advanc chang queri queri addit applic requir fast approxim answer rknn queri scenario import gener fast answer high recal paper propos approach effici approxim rknn search arbitrari metric space queri time approach advantag exist metric index structur propos approxim nearest neighbor distanc order prune search space show method scale significantli exist approxim approach produc approxim true queri result high recal detect visual subspac cluster hierarchi abstract subspac cluster call project cluster address problem set attribut relev cluster high dimension featur space paper propos algorithm dish detect subspac cluster hierarchi improv point exist approach dish detect cluster subspac significantli dimension dish uncov complex hierarchi nest subspac cluster cluster lower dimension subspac embed higher dimension subspac cluster hierarchi consist singl inclus exhibit multipl inclus model graph tree dish detect cluster size shape densiti propos visual complex hierarchi mean visual model call subspac cluster graph relationship subspac cluster explor glanc compar experi show perform effect dish interv dfocus similar search time seri databas proxim queri larg traffic network paper present origin network graph embed speed distanc rang nearest neighbor queri weight graph approach implement paradigm filter refin queri process proxim queri static dynam object present embed comput lower upper bound filter distanc approxim true shortest path distanc significantli tradit filter euclidean distanc distanc approxim filter step prune true drop true hit refin step order guid inform search experiment evalu real world data set demonstr signific perform boost propos concept exist work robust complet effici correl cluster correl cluster aim detect data point hyperplan data space exhibit common correl subset featur recent propos method correl cluster suffer sever drawback includ poor robust nois paramet set incomplet result miss cluster poor usabl due complex input paramet poor scalabl paper propos correl cluster algorithm copac correl partit cluster aim improv robust complet usabl effici experiment evalu empir show copac superior exist state art correl cluster method term runtim accuraci complet result explor complex relationship correl cluster high dimension data cluster exist arbitrarili orient subspac featur space addit call correl cluster complex relationship correl cluster subspac form line enclos correl cluster superspac form plane gener relationship complex hierarchi multipl inclus cluster embed super cluster uncov hierarch relationship detect correl cluster import inform gain exist approach detect complex hierarch relationship correl cluster propos algorithm eric tackl problem visual result mean graph base represent experiment evalu show eric find inform state art correl cluster method outperform exist competitor term effici gener optim multi dstep dnearest neighbor queri process abstract similar search algorithm directli reli index structur requir lot distanc comput applic databas complex object defin costli distanc function spatial tempor multimedia data adequ multi step queri process strategi crucial perform similar search routin deal complex distanc function reduc number candid return filter step evalu refin step fundament effici queri process state art multi step nearest neighbor knn search algorithm design lower bound distanc estim candid prune applic upper bound distanc approxim addition reduc number candid paper gener tradit concept optim introduc notion ri optim depend distanc inform filter step propos multi step knn search algorithm util lower upper bound distanc inform ilu filter step show contrast exist approach propos solut ri lu optim experiment evalu demonstr signific perform gain exist method futur trend data mine knowledg discoveri dmkd held annual conjunct acm sigmod confer workshop aim bring data mine research experienc practition goal discuss gener data mine tool miniconfer format focus present polish research result dmkd workshop aim foster inform atmospher research practition freeli interact short present open discuss ongo work forward research vision experi futur data mine applic addit research data mine algorithm experi innov mine system applic interest year dmkd workshop broad theme futur trend data mine topic interest includ privaci amp secur issu mine privaci preserv data mine techniqu invalu case detail data allow infer high level inform relev mine nation secur applic mine data stream emerg applic data arriv process continu basi mine benefit pass static persist snapshot data mine bioinformat biolog databas system high perform data mine tool plai crucial role analysi grow databas bio sequenc structur semi dsupervis threshold queri pharmacogenom time sequenc analysi time seri data capit import pharmacogenom experiment evalu base observ time depend reaction behavior organ data mine time seri databas import instrument understand effect drug individu complex natur time seri pose big challeng effect effici data mine paper focu detect tempor depend time seri introduc analysi concept threshold queri semi supervis extens support paramet set appli train dataset basic threshold queri report time seri exceed user defin queri threshold time frame semi supervis threshold queri threshold automat adjust characterist data set train dataset order support threshold queri effici present effici access method fact partial inform time seri requir queri time extens experiment evalu demonstr perform solut show semi supervis threshold queri appli gene express data worthwhil approxim revers dnearest neighbor queri gener metric space paper propos approach effici approxim rknn search arbitrari metric space queri time method approxim nearest neighbor distanc order prune search space experi solut scale significantli exist approxim approach produc approxim true queri result high recal similar search time seri base threshold queri abstract similar search time seri data requir applic field promin work focus similar search complet time seri similar subsequ time seri domain financi analysi medicin environment meteorolog environment observ detect tempor depend time seri import contrast tradit approach time seri purpos match coars trend inform time seri suffici solv mention problem tempor depend time seri detect determin point time time seri exce specif threshold paper introduc concept threshold queri time seri databas report time seri exceed user defin queri threshold similar time frame compar queri time seri present effici access method fact partial inform time seri requir queri time perform solut demonstr extens experiment evalu real world artifici time seri data tquest threshold queri execut larg set time seri similar search time seri base threshold queri abstract similar search time seri data requir applic field promin work focus similar search complet time seri similar subsequ time seri domain financi analysi medicin environment meteorolog environment observ detect tempor depend time seri import contrast tradit approach time seri purpos match coars trend inform time seri suffici solv mention problem tempor depend time seri detect determin point time time seri exce specif threshold paper introduc concept threshold queri time seri databas report time seri exceed user defin queri threshold similar time frame compar queri time seri present effici access method fact partial inform time seri requir queri time perform solut demonstr extens experiment evalu real world artifici time seri data threshold similar queri larg time seri databas similar search time seri data activ area research paper introduc concept threshold similar queri time seri databas report time seri exceed user defin queri threshold similar time frame compar queri time seri addit present data structur support threshold similar queri effici perform solut demonstr extens experiment evalu deriv quantit model correl cluster correl cluster aim group data set correl cluster object cluster exhibit densiti common arbitrarili orient hyperplan arbitrari dimension algorithm task propos recent algorithm comput partit data cluster step pipelin advanc data analysi system model post cluster step deriv quantit model correl cluster address paper describ origin approach handl step introduc gener method extract quantit inform linear depend correl cluster concept independ cluster model appli post process step correl cluster algorithm show quantit model predict probabl distribut object creat model broad experiment evalu demonstr benefici impact method applic signific practic import deli dclu boost robust find hierarchi subspac cluster abstract cluster algorithm applic high dimension featur space cluster exist specif subspac origin featur space cluster call subspac cluster paper propos algorithm hisc hierarch subspac cluster detect hierarchi nest subspac cluster relationship lowerdimension subspac cluster embed higher dimension subspac cluster compar experi synthet real data set show perform effect hisc effici revers dnearest neighbor search arbitrari metric space revers nearest neighbor rknn problem find object data set nearest neighbor includ queri object gener revers nearest neighbor problem receiv increas attent recent industri scientif applic call solut rknn problem arbitrari metric space data object euclidean metric distanc function object similar applic solut gener problem advanc chang queri queri exist approach design specif nn problem addit knowledg previous propos method gener rknn search applic euclidean vector data gener metric object paper propos approach effici rknn search arbitrari metric space queri time approach advantag exist metric index structur propos conserv progress distanc approxim order filter true drop true hit approxim nearest neighbor distanc data object upper lower bound function paramet method gener consider storag overhead show broad experiment evalu real world data scalabl usabl approach mine hierarchi correl cluster detect correl featur high dimension data set import data mine task correl arbitrarili complex featur correl featur nois featur actual depend cluster cluster point locat common hyperplan arbitrari dimension data space gener separ arbitrarili orient subspac origin data space recent propos algorithm design uncov correl cluster disadvantag method detect correl cluster dimension nest complet hierarch structur correl cluster vari dimension detect hierarch cluster approach propos algorithm hico hierarch correl order hierarch approach correl cluster algorithm determin cluster hierarchi visual correl diagram compar experi synthet real data set show perform effect hico effici queri process arbitrari subspac vector approxim paper introduc partial vector approxim file extens vector approxim file construct effici answer partial similar queri subspac idea partial va file divid va file separ file dimens load dimens answer queri partial va file construct improv queri perform system cope wide varieti previous unknown queri subspac propos algorithm partial knn rang queri base partial va file experi demonstr propos partial va file algorithm improv averag queri perform comparison origin va file answer partial similar queri time seri analysi concept adapt threshold similar issu data mine time seri databas utmost import practic applic attract lot research past year paper focu recent propos concept threshold similar compar time seri base time frame exce user defin amplitud threshold propos approach cluster analysi time seri base adapt threshold similar import issu threshold similar choic threshold threshold automat adapt characterist small train dataset concept support vector machin optim learn small train set order yield accur cluster entir time seri databas experiment evalu demonstr cluster analysi adapt threshold similar successfulli appli scientif real world data mine applic onlin hierarch cluster data warehous environ import industri applic reli data mine method uncov pattern trend larg data warehous environ data warehous typic updat period batch mode mine pattern updat requir accuraci data mine method fast avail date knowledg presenc heavi updat load cope problem propos onlin data mine algorithm perman store discov knowledg suitabl data structur enabl effici adapt structur insert delet raw data paper demonstr hierarch cluster method reformul onlin algorithm base hierarch cluster method optic densiti estim data group discuss algorithm schema special effici onlin singl link cluster broad experiment evalu demonstr effici superior signific speed factor larg bulk insert delet gener framework effici subspac cluster high ddimension data subspac cluster investig extens tradit cluster algorithm fail detect meaning cluster high dimension data space recent propos subspac cluster method suffer sever problem algorithm typic scale exponenti data dimension subspac dimension cluster perform reason algorithm global densiti threshold cluster question cluster subspac significantli dimension exhibt significantli vari densiti paper propos gener framework overcom limit framework base effici filterrefin architectur scale quadrat data dimension dimension subspac cluster appli cluster notion includ notion base local densiti threshold broad experiment evalu synthet real world data empir show method achiev signific gain runtim qualiti comparison state art subspac cluster algorithm effect effici distribut model dbase cluster select estim high dimension window queri cluster accur effici similar search object point sampl abstract fast evolv resourc object protein data bank pdb world wide web techniqu call similar model effici effect search object indispens invari specif geometr transform scale translat rotat import featur similar model paper focu rotat invari propos method repres object accur context rotat invari voxel techniqu addit extend exist featur base similar model propos spheric partit data object base proportion redund gener exist method featur extract broad experiment evalu compar method exist method term accuraci effici experiment confirm point sampl method suit repres object context rotat invari voxel represent addit empir show similar model significantli outperfom competit rotat invari model term accuraci effici boss brows optic dplot similar search similar search databas system increasingli import task modern applic domain multimedia molecular biologi medic imag comput aid engin market purchas assist paper show visual hierarch cluster structur databas object aid user time consum task find similar object present relat work explain shortcom led develop method base reachabl plot introduc approach automat extract signific cluster hierarch cluster represent suitabl cluster repres techniqu basi visual data mine implement algorithm result industri prototyp experiment evalu evalu base real world test data set point approach automat cluster recognit extract cluster repres creat meaning result compar short time subspac select cluster high ddimension data high dimension featur space tradit cluster algorithm tend break term effici qualiti data set cluster hidden subspac origin featur space paper present featur select techniqu call surf subspac relev cluster find subspac interest cluster sort relev sort base qualiti criterion interesting subspac nearest neighbor distanc object method parameterless address unsupervis notion data mine task cluster broad evalu base synthet real world data set demonstr surf suitabl find relev subspac high dimension spars data set produc result compar method densiti connect cluster local subspac prefer cluster algorithm tend break high dimension featur space cluster exist specif subspac attribut subset origin featur space task project cluster subspac cluster defin recent solut tackl problem propos concept local subspac prefer captur main direct high point densiti concept adopt densiti base cluster cope high dimension data achiev advantag exist approach propos method determin result depend order process robust nois perform singl scan databas linear number dimens broad experiment evalu show approach yield result significantli qualiti recent work cluster high dimension data visual mine cluster hierarchi similar search databas system increasingli import task modern applic domain multimedia molecular biologi medic imag comput aid engin market purchas assist paper show visual hierarch cluster structur databas object aid user time consum task find similar object present relat work explain shortcom led develop method base reachabl plot introduc approach automat extract signific cluster hierarch cluster represent suitabl cluster repres techniqu basi visual data mine implement algorithm result industri prototyp experiment evalu evalu base real world test data set point approach automat cluster recognit extract cluster repres creat meaning result compar short time densiti dconnect subspac cluster high ddimension data applic domain molecular biologi geographi produc tremend amount data longer manag effici effect data mine method primari data mine task cluster tradit cluster algorithm fail detect meaning cluster real world data set character high dimension inher spars data space data set interest cluster hidden subspac origin featur space concept subspac cluster recent address aim automat identifi subspac featur space cluster exist paper introduc subclu densiti connect subspac cluster effect effici approach subspac cluster problem concept densiti connect underli algorithm dbscan eksx subclu base formal cluster notion contrast exist grid base approach subclu detect arbitrarili shape posit cluster subspac monoton densiti connect effici prune subspac process gener cluster bottom examin unnecessari subspac subclu deliv subspac cluster dbscan found appli subspac separ support vector machin classifi larg set multi drepres object databas kei technolog molecular biologi data intens disciplin molecular biolog databas heterogen unif data integr mandatori make huge amount inform promis approach integr ontolog map biolog entiti ontolog achiev manual semiautomat system automat classif biolog entiti ontolog save time effort present support vector machin base approach automat classifi biolog entiti ontolog solv difficult task method cope aspect biolog entiti belong class class vari abstract level object represent classifi enabl draw inform possibl object incomplet method introduc techniqu objectadjust weight regul impact represent dynam object significantli improv time perform classifi exploit inherit relat ontolog experiment evalu protein data part establish molecular biolog ontolog show prototyp offer impress accuraci effici cope larg number class encount real world problem support german ministeri educ scienc visual mine cluster hierarchi similar search databas system increasingli import task modern applic domain multimedia molecular biologi medic imag comput aid engin market purchas assist paper show visual hierarch cluster structur databas object aid user time consum task find similar object present relat work explain shortcom led develop method base reachabl plot introduc approach automat extract signific cluster hierarch cluster represent suitabl cluster repres techniqu basi visual data mine implement algorithm result industri prototyp experiment evalu evalu base real world test data set point approach automat cluster recognit extract cluster repres creat meaning result compar short time immer gr dfere und komplexer datenmengen herausforderungen fcr cluster dalgorithmen bioinformat databas state art research perspect bioinformat comput biologi applic mathemat comput scienc method solv problem molecular biologi requir larg scale data comput analysi research area receiv consider attent databas plai essenti role molecular biologi bioinformat molecular biologi data cheap produc lead prolifer databas number bioinformat databas access worldwid li molecular biologi data molecular biologi literatur literatur refer store databas bioinformat databas larg sequenc databas genbank nucleotid sequenc gener grow rapidli abstract ad month literatur databas pubm bioinformat databas heterogen data data model paradigm manag system data analysi tool support bioinformat databas implement queri updat manag method rare appli databas present aim introduc current bioinformat databas stress aspect depart convent databas detail survei found effect similar search voxel cad object similar search databas system increasingli import task modern applic domain multimedia molecular biologi medic imag cad applic suitabl similar model clear represent result reduc cost develop produc part maxim reus exist part paper adapt similar model voxel cad data introduc model base eigenvector experiment evalu similar model base real world test dataset introduc hierarch cluster effect analys compar similar model show similar model evalu procedur suitabl industri rank interest subspac cluster high dimension data applic domain life scienc molecular biologi produc tremend amount data longer manag cient ectiv data mine method primari data mine task cluster tradit cluster algorithm fail detect meaning cluster high dimension inher spars featur space real world data set data set cluster hidden subspac origin featur space present pre process step tradit cluster algorithm detect interest subspac high dimension data cluster purpos defin qualiti criterion interesting subspac propos cient algorithm call ri rank nterest subspac examin subspac broad evalu base synthet real world data set empir show ri suitabl find relev subspac larg high dimension spars data rank increment optic effici comput updat hierarch cluster order data warehous challeng field applic data mine task cluster updat collect appli data warehous period batch mode consequ mine pattern discov data warehous cluster structur updat paper present method increment updat cluster structur comput hierarch cluster algorithm optic determin part cluster order affect updat oper develop effici algorithm increment updat exist cluster order rank interest subspac cluster high dimension data applic domain life scienc molecular biologi produc tremend amount data longer manag cient ectiv data mine method primari data mine task cluster tradit cluster algorithm fail detect meaning cluster high dimension inher spars featur space real world data set data set cluster hidden subspac origin featur space present pre process step tradit cluster algorithm detect interest subspac high dimension data cluster purpos defin qualiti criterion interesting subspac propos cient algorithm call ri rank nterest subspac examin subspac broad evalu base synthet real world data set empir show ri suitabl find relev subspac larg high dimension spars data rank set featur vector similar search voxel cad object modern applic domain multimedia molecular biologi medic imag similar search databas system increasingli import task cad applic suitabl similar model reduc cost develop produc part maxim reus exist part exist similar model base featur vector paper shortli review model pursu paradigm base promis model explain set featur vector ectiv cient similar search introduc intuit distanc measur set featur vector algorithm cient comput present method acceler process similar queri vector set data experiment evalu base real world test data set point similar approach yield meaning result compar short time comput biologi databas digest data comput biologi bioinformat defin applic mathemat comput scienc method solv problem molecular biologi requir larg scale data comput analysi expect molecular biologi databas plai essenti role comput biologi research develop paper introduc current molecular biologi databas stress data model data acquisit data retriev integr molecular biologi data sourc paper primarili intend audienc comput scientist limit background biologi datenbanken der bioinformatik data bubbl qualiti preserv perform boost hierarch cluster paper investig scale hierarch cluster method optic extrem larg databas util data compress method birch random sampl propos step procedur compress data suitabl repres object appli hierarch cluster algorithm object recov cluster structur data set base result compress data kei issu approach design compress data item hierarch cluster algorithm appli inform infer cluster structur origin data set step crucial result hierarch cluster algorithm appli naiv random sampl cluster featur cf gener birch deterior rapidli higher compress rate due kei problem identifi solv problem propos effici post process step concept data bubbl special kind compress data item appli optic data bubbl recov accur approxim cluster structur larg data set high compress rate comprehens perform qualiti evalu show trade qualiti cluster result great increas perform 