rank set search medlin document topic evolut stream document visual cluster result hierarch cluster long histori numer taxonomi machin learn applic field data mine statist analysi inform retriev cluster involv find specif number subgroup set observ data point object lwa sponsor lwa lernen wissen adapt sponsor fast approxim duplic detect dnmr spectra evalu text retriev method similar search multi ddimension nmr dspectra abstract search mine nuclear magnet reson nmr spectra natur occur substanc import task investig potenti chemic compound multi dimension nmr spectra relat object document consist continu multi dimension point call peak word develop map continu nmr spectra discret textlik data map text retriev method appli evalu perform retriev method standard vector space model probabilist latent semant index plsi plsi learn hidden topic data case nmr data interest own right addition develop evalu simpl direct similar function detect duplic nmr spectra experi show vector space model plsi design text data creat human effect handl map nmr data origin natur product addition plsi find meaning topic nmr data fast approxim duplic detect dnmr spectra applic latent topic document analysi larg dscale proteom databas bayesian fold din dirichlet kernel plsi denclu fast cluster base kernel densiti estim denclu algorithm emploi cluster model base kernel densiti estim cluster defin local maximum estim densiti function data point assign cluster hill climb point local maximum put cluster disadvantag denclu hill climb make unnecessari small step begin converg maximum close introduc hill climb procedur gaussian kernel adjust step size automat extra cost prove procedur converg local maximum reduc special case expect maxim algorithm show experiment procedur iter acceler sampl base method sacrif small amount accuraci cluster em dbase hill dclimb kernel densiti estim sponsor workshop knowledg discoveri combin multipl classifi interest multimedia system data type modal mine analyz wearabl experi collect system unobtrus record wearer convers recogn face dialog partner rememb voic system see person face hear voic summari convers person remind wearer correctli identifi person mixtur video audio stream classif judgment individu modal classifi combin effect yield accur decis address problem combin strategi previou studi meta classif strategi support vector machin propos preliminari result show combin face recognit speaker identif technolog metaclassif dramat effect weight interpol meta classif gener appli applic combin multipl classifi modif cluster em dbase hill dclimb kernel densiti estim sponsor analys latent topic larg dscale proteom sponsor analys latent topic larg dscale proteom sponsor duplic detect dnmr spectra nuclear magnet reson nmr spectra structur analysi small molecul contrast nmr spectra nmr spectra correl chemic shift time spectrum consist peak twodimension space import inform peak locat center captur bond relationship hydrogen carbon atom spectrum inform chemic structur product case structur read simpl straightforward manner structur elucid involv consider amount manual effort high field nmr spectromet nmr spectra record short time common situat lab compani repositori nmr spectra partial annot structur inform remain spectra structur unknown case research lab collabor repositori merg annot share reduc problem task find duplic set nmr spectra propos simpl robust definit nmr duplic small measur error give quadrat algorithm problem implement sql analyz abstract class heurist base select peak heurist work filter step pair duplic fals posit compar method respect run time final discuss effect duplic definit real data tagungsband zum gi dworkshop fcber grundlagen von datenbanken gi dworkshop foundat databas evalu text retriev method similar search multi ddimension nmr dspectra abstract search mine nuclear magnet reson nmr spectra natur occur substanc import task investig potenti chemic compound multi dimension nmr spectra relat object document consist continu multi dimension point call peak word develop map continu nmr spectra discret textlik data map text retriev method appli evalu perform retriev method standard vector space model probabilist latent semant index plsi plsi learn hidden topic data case nmr data interest own right addition develop evalu simpl direct similar function detect duplic nmr spectra experi show vector space model plsi design text data creat human effect handl map nmr data origin natur product addition plsi find meaning topic nmr data similar search multi ddimension nmr dspectra natur product abstract search mine nuclear magnet reson nmr spectra natur occur product import task investig potenti chemic compound develop set base similar function suffici captur abstract aspect similar nmr spectra document consist continu multi dimension point word probabilist semant index plsi retriev method learn hidden topic develop map continu nmr spectra discret text data map includ redund discret data proof help plsi model experi show plsi design text data creat human effect handl map nmr data origin natur product addition plsi combin map find meaning topic nmr data dimens induc cluster optim multipl top dk queri join advanc data mine applic requir support relat databas engin cluster applic high dimension featur space demand proper support multipl top queri order perform project cluster research tackl problem optim restrict rank top queri solut singl rank criterion deficit optim multipl topk queri join target paper perspect hand propos minim handi extens sql express multipl top queri hand propos optim hash join strategi effici execut type queri extens experi conduct context show feasibl propos multiresolut similar search imag databas hnlichkeitssuch musik ddatenbanken mit hilf von visualisierungen visual interact solv complex optim problem visual interact solv complex optim problem hd dey visual cluster high dimension data databas support dprotein data set analysi combi doper databas support data mine applic databas support data mine import research topic larg high dimension data volum comprehens support databas side paper identifi data intens subproblem aggreg high dimension data low dimension project instanc estim low dimension histogram occur establish data mine techniqu show exist olap sql extens insu cient high dimension data propos sql oper seamlessli fit set exist olap group oper gener approach cluster larg databas nois multi dresolut similar search imag databas hd dey visual cluster high dimension data surpris behavior distanc metric high dimension space recent year effect curs high dimension studi great detail problem cluster nearest neighbor search index high dimension space data spars tradit index algorithm techniqu fail effici effect perspect recent research result show high dimension space concept proxim distanc nearest neighbor qualit meaning paper view dimension curs point view distanc metric measur similar object specif examin behavior commonli lk norm show problem meaning high dimension sensit mean manhattan distanc metric norm consist prefer euclidean distanc metric norm high dimension data mine applic intuit deriv analysi introduc examin natur extens lk norm fraction distanc metric show fraction distanc metric meaning result theoret empir perspect result show fraction distanc metric significantli improv effect standard cluster algorithm mean algorithm cluster dstructur small aminoacid dchain detect depend sequenti context protein high dimension cluster larg data set nearest neighbor high dimension space nearest neighbor search high dimension space interest import problem relev wide varieti databas applic recent result show problem difficult perform issu qualiti issu paper discuss qualiti issu identifi gener notion nearest neighbor search relev problem high dimension space contrast previou approach notion nearest neighbor search treat dimens equal qualiti criterion select relev dimens project respect queri qualiti criterion rate data cluster queri point select project propos effici effect algorithm solv gener nearest neighbor problem experi base number real synthet data set show approach insight natur nearest neighbor search high dimension data cluster techniqu larg data set past futur interest kdd dnew sigmod cluster techniqu larg data set past futur cluster method larg databas past futur cluster algorithm attract task class identif spatial databas applic larg spatial databas rise requir cluster algorithm minim requir domain knowledg determin input paramet discoveri cluster arbitrari shape good effici larg databas cluster algorithm offer solut combin requir paper present cluster algorithm dbscan reli densiti base notion cluster design discov cluster arbitrari shape dbscan requir input paramet support user determin perform experiment evalu effect effici dbscan synthet data real data sequoia benchmark result experi demonstr dbscan significantli effect discov cluster arbitrari shape algorithm clar an dbscan outperform claran factor term effici cluster method larg databas past futur cluster algorithm attract task class identif spatial databas applic larg spatial databas rise requir cluster algorithm minim requir domain knowledg determin input paramet discoveri cluster arbitrari shape good effici larg databas cluster algorithm offer solut combin requir paper present cluster algorithm dbscan reli densiti base notion cluster design discov cluster arbitrari shape dbscan requir input paramet support user determin perform experiment evalu effect effici dbscan synthet data real data sequoia benchmark result experi demonstr dbscan significantli effect discov cluster arbitrari shape algorithm clar an dbscan outperform claran factor term effici optim grid dcluster break curs dimension high ddimension cluster applic requir cluster larg amount high dimension data cluster algorithm work effect effici high dimension space due call curs dimension addit high dimension data signific amount nois addit effect problem paper review compar exist algorithm cluster high dimension data show impact curs dimension effect effici comparison reveal condens base approach birch sting promis candid achiev effici show basic condens base approach sever weak respect effect high dimension space overcom problem develop cluster techniqu call optigrid base construct optim grid hd dey visual mine high ddimension data introduct motiv visual enabl clear overview detail associ element larg categor data set approach address problem produc limit result detect relationship amino acid residu sequenc protein structur base class rotam librari past good number rotam librari publish deeper understand conform behaviour amino acid residu protein number high resolut rai protein structur grown significantli year comprehens analysi conform behaviour todai paper present method compil class rotam librari detect interest relationship residu conform sequenti context protein method base algorithm cluster residu conform demonstr effect method appli algorithm librari consist tripetid fragment form nativ amino acid analysi show interest result specif tripeptid fragment show unexpect conform residu highli prefer conform neigh effici approach cluster larg multimedia databas nois cluster algorithm appli cluster larg multimedia databas effect effici exist algorithm limit cluster multimedia databas requir cluster high dimension featur vector multimedia databas larg amount nois paper introduc algorithm cluster larg multimedia databas call denclu densitybas cluster basic idea approachi model point densiti analyt sum influenc function data point cluster identifi determin densiti attractor cluster arbitrari shape easili simpl equat base densiti function advantag approach firm mathemat basi good cluster properti data set larg amount nois compact mathemat 