comput intellig method bioinformat biostatist comput intellig bioengin essai memori antonina starita comput intellig method bioinformat biostatist approach biclust gene express data modular singular decomposit approach biclust gene express data modular singular decomposit search microrna prostat cancer target gene experiment valid index fuzzi cluster similar experiment valid index fuzzi discrimin power random featur comput intellig machin learn bioinformat natur comput method bioinformat data analysi problem bioinformat concern fusion multisensor output fusion multi sourc inform integr kind biolog data natur comput possibl bioinformat present interest natur inspir methodolog handl complex problem articl survei role natur comput domain protein structur predict microarrai data analysi gene regulatori network gener util learn abil neural network adapt uncertainti handl capac fuzzi set rough set model ambigu search potenti genet algorithm effici travers larg search space natur comput method bioinformat survei data analysi problem bioinformat concern fusion multisensor output fusion multi sourc inform integr kind biolog data natur comput possibl bioinformat present interest natur inspir methodolog handl complex problem articl survei role natur comput domain protein structur predict microarrai data analysi gene regulatori network gener util learn abil neural network adapt uncertainti handl capac fuzzi set rough set model ambigu search potenti genet algorithm effici travers larg search space soft rank cluster due diffus larg dimension data set dna microarrai document organ retriev applic grow interest cluster method base proxim matrix advantag base data structur size depend cardin dimension paper propos cluster techniqu base fuzzi rank rank help overcom issu larg dimension data set fuzzi formul encod inform contain smallest entri proxim matrix compar experi present standard hierarch cluster techniqu refer kei word fuzzi rank cluster data mine dna microarrai stabil perform biclust algorithm abstract stabil import properti machin learn algorithm stabil cluster relat cluster qualiti ensembl divers wai achiev deeper understand confid bioinformat data analysi specif field fuzzi biclust stabil analyz port definit exist stabil index fuzzi set adapt biclust problem paper present work direct select repres stabil index experiment verifi compar properti experiment result present gener agreement differ select method experiment comparison kernel cluster method abstract paper compar perform popular kernel cluster method data set method base central cluster incorpor wai concept fuzzi cluster kernel machin data set sampl applic domain size discuss techniqu valid result present result cluster kernel space gener outperform standard cluster method proven consist marketa zvelebil jeremi baum survei kernel spectral method cluster applic fuzzi set theori abstract paper present short survei fuzzi neural approach inform retriev goal approach de ne exibl inform retriev system deal inher vagu uncertainti retriev process survei address approach met goal local learn tide level time seri fuzzi approach membership embed space approach spectral cluster possibilist cluster featur space abstract paper propos possibilist mean featur linear fuzzi cluster select variabl grade possibilist approach vector quantiz fuzzi rank imag reconstruct fuzzi logic applic basic concept fuzzi system point view fuzzi logic base control chapter state art fuzzi control power system outlin support bibliographi literatur area applic fuzzi logic control power system stabil illustr exampl fairli straightforward applic fuzzi logic learn algorithm possibilist approach biclust applic oligonucleotid microarrai data analysi abstract import research object identifi gene similar behavior respect condit recent tackl biclust techniqu paper introduc approach biclust problem possibilist cluster paradigm propos possibilist biclust algorithm find biclust time assign membership biclust gene condit biclust problem maxim size biclust minim residu face optim proper function appli algorithm yeast databas obtain fast converg good qualiti solut discuss effect paramet tune sensit method paramet valu comparison method literatur present supervis classif gene select simul anneal formul grade possibilist approach fuzzi cluster abstract fuzzi cluster tool captur intrins structur data set paper propos formul soft transit fuzzi membership probabilist partit possibilist propos techniqu free membership introduc addit penalti term possibilist mean featur propos techniqu demonstr numer experi soft transit probabilist possibilist fuzzi cluster discuss grade possibilist model review cluster algorithm deriv basic mean introduc formal provid altern unifi perspect cluster algorithm focus membership cost function interest case concept grade possibl formul includ extrem case probabilist assumpt possibilist assumpt formul state interv equal constraint enforc normal condit requir grade possibilist condit outlin basic grade possibilist cluster algorithm experiment demonstr present aim highlight properti attain implement suitabl grade possibilist model advanc fuzzi set rough set possibilist cluster approach trackless ring pattern recognit rich counter share farthest neighbor approach cluster high dimension cluster algorithm routin biomed disciplin basic tool bioinformat depend task hand popular option central partit techniqu agglom hierarch cluster techniqu deriv method studi establish categori drawback relat data dimension partit algorithm bottom structur hierarch agglom algorithm overcom limit motiv problem gene express analysi dna microarrai present hierarch cluster algorithm base complet principl analysi share farthest neighbor present framework cluster rank index introduc share farthest neighbor cluster criterion illustr properti method present experiment result data set strategi evalu data cluster extrins knowledg class label unsupervis gene select cluster simul anneal soft rank cluster due diffus larg dimension data set dna microarrai document organ retriev applic grow interest cluster method base proxim matrix advantag base data structur size depend cardin dimension paper propos cluster techniqu base fuzzi rank rank help overcom issu larg dimension data set fuzzi formul encod inform contain smallest entri proxim matrix compar experi present standard hierarch cluster techniqu refer kei word fuzzi rank cluster data mine dna microarrai fuzzi approach imag analysi hla type oligonucleotid microarrai integr conceptu densiti wordnet domain cald gloss noun sens disambigu experiment analysi depend codeword bit error ecoc learn machin main factor ect ectiv error correct output code ecoc method classif depend error comput codeword bit present extens experiment work evalu depend output error decomposit unit ecoc learn machin appli measur base mutual inform compar depend ecoc multi layer perceptron ecoc mlp made singl multi input multi output mlp ecoc ensembl made set independ parallel dichotom ecoc pnd experiment analyz relationship architectur depend output error perform ecoc learn machin result show depend comput codeword bit significantli smaller ecoc pnd point ensembl independ parallel dichotom suit implement ecoc classif method experiment result suggest architectur ecoc learn machin improv independ output error divers base learner effect error correct output code method ensembl monolith learn machin error correct output code ecoc method multiclass classif present open problem rang trade error recov capabl learnabl induc dichotomi select proper base learner design separ code multiclass problem experiment analyz main factor affect effect ecoc method show architectur ecoc learn machin influenc accuraci ecoc classifi highlight ensembl parallel independ dichotom multi layer perceptron suit implement ecoc method quantit evalu depend codeword bit error mutual inform base measur experiment show low depend enhanc gener capabl ecoc show proper select base learner decod function reconstruct stage significantli affect perform ecoc ensembl analysi relationship error recov power accuraci base learner depend codeword bit show factor concur effect ecoc method straightforward depend distribut complex data automat noun sens disambigu abstract paper explor fulli automat knowledg base method perform noun sens disambigu reli wordnet ontolog basi method idea conceptu densiti correl sens word context formula calcul conceptu densiti propos evalu semcor corpu extens conceptu densiti task word sens disambigu wsd consist examin word token sens word wordnet wn ontolog base synset set synonym extern lexic resourc perform wsd task wsd approach word disambigu portion text embed context initi input sourc inform word context process lexic knowledg sourc wn fulli automat method requir kind train process need perform wsd conceptu densiti cd measur correl sens word context foundat measur conceptu distanc defin length shortest path connect concept hierarch semant net start point work cd formula agirr rigau compar area subhierarchi cd pm gamma nhyp fuzzi concept vector quantiz train gene select random voronoi ensembl algorithm model paradigm shift fuzzi cluster applic ensembl techniqu base singular spectrum analysi daili rainfal forecast random voronoi ensembl gene select paper address issu assess import input variabl respect dichotom classif problem linear linear case consid linear case applic deriv base salienc yield commonli adopt rank criterion linear case method extend introduc resampl techniqu cluster obtain result stabil estim todo autom tico para la desambiguaci xica de nombr boost classif electron nose data ensembl learn machin ensembl learn machin constitut main current direct machin learn research appli wide rang real problem absenc unifi theori ensembl theoret reason combin multipl learner empir evid effect approach paper present overview ensembl method explain main reason outperform singl classifi ensembl propos taxonomi base main wai base classifi gener combin theoret experiment analysi dstage system classif abstract popular approach multicategori classif task stage system base global classifi reject local nearest neighbor classifi pattern reject classifi classifi output reject pattern pass nearest neighbor classifi top rank class return classifi nearest neighbor classifi pattern top class classifi reject pattern edit strategi nearest neighbor refer databas control classifi consid analyz system show level nearest neighbor classifi optim bay sens system optim formal relat respons time system reject rate classifi system paramet error respons time trade discuss final experiment studi instanc system appli recognit handwritten digit system classifi fuzzi basi function network system feed forward neural network classif result respons time set system paramet report system index term multicategori classif reject global local classif hierarch classifi bay classifi neurobject object dorient librari neural network develop neurobject set librari class neural network develop exploit potenti object orient design program main goal librari consist support experiment research neural network fast prototyp induct machin learn applic present neurobject design issu main function program exampl show map neural network concept design librari class possibilist ring detect rich pattern recognit paper present result obtain algorithm core possibilist spheric shell algorithm detect complex imag full ring preliminari knowledg number posit ring algorithm depend choic critic paramet statist studi paper lead high success rate larg number ring depend codeword bit error ecoc learn machin experiment analysi parallel linear dichotom present learn machin model classic problem base decomposit multiclass classic problem set class subproblem assign linear dichotom learn task independ experiment perform classic data set show learn machin model achiev signic perform improv mlp previou classier model base decomposit polychotomi dichotomi theoret reason good properti gener propos learn machin model explain framework statist learn theori keyword gener learn machin classic decomposit polychotomi dichotomi statist learn theori introduct learn method implement induct principl empir risk minim regular structur risk minim bayesian infer mi compar decomposit method classif decomposit method multiclass classic problem constitut power framework improv gener capabl larg set learn machin includ support vector machin multi layer perceptron present review main decomposit approach classi cation experiment comparison onep class opc correct classier cc error correct output code ecoc decomposit method implement multi layer perceptron dichotom result show cc ecoc outperform opc consid data set introduct decomposit method classic split multiclass problem polychotomi seri independ twoclass problem dichotomi recompos output dichotom order reconstruct origin polychotomi learn machin implement decomposit method compos main unit decomposit unit analyz input pa pattern recognit rich counter possibilist dspheric shell algorithm effect error correct output code multiclass learn problem framework decomposit method multiclass classifi problem error correct output code ecoc fruitfulli codeword code class order enhanc gener capabl learn machin effect error correct output code depend independ codeword bit accuraci dichotomi learn separ linear dichotom improv independ comput codeword bit fulli exploit error recov capabl ecoc experiment present paper compar ecoc decomposit method implement monolith multi layer perceptron set linear linear independ dichotom effect ecoc decomposit scheme obtain parallel linear dichotom pnd learn machin base decomposit polychotomi dichotomi linear independ dich proceed icsc symposia intellig industri autom iia soft comput soco neurobject set librari class neural network develop neurobject set librari class neural network develop exploit potenti object orient design program main goal librari support fast prototyp induct machin learn applic base neural network paper present librari design issu main function simpl exampl program neurobject introduct neural network plai import role machin learn permitt effic face problem regress classif neural network relev compon complex system induct learn task nowadai limit diffus neural network technolog industri applic depend high cost relat long develop time neural network algorithm implement scratch order emb tool softwar product simul anneal dmean cluster algorithm high speed industri neural network classif system base custom vlsi implement neurobject set librari class neural network develop neurobject set librari class neural network develop exploit potenti object orient design program main goal librari support fast prototyp induct machin learn applic base neural network paper present librari design issu main function simpl exampl program neurobject introduct neural network plai import role machin learn permitt effic face problem regress classif neural network relev compon complex system induct learn task nowadai limit diffus neural network technolog industri applic depend high cost relat long develop time neural network algorithm implement scratch order emb tool softwar product reliabl neuro dcontrol arrai sensor cross dmonitor work dstate fuzzi cluster base segment system support diagnosi medic imag 