Where can buy Intelligent data engineering and automated learning ideal 2014 15th international conf

Page 1


Intelligent Data Engineering and Automated Learning IDEAL

2014 15th International Conference Salamanca

Spain September 10 12 2014

Proceedings 1st Edition Emilio Corchado

Visit to download the full and correct content document: https://textbookfull.com/product/intelligent-data-engineering-and-automated-learning-i deal-2014-15th-international-conference-salamanca-spain-september-10-12-2014-pro ceedings-1st-edition-emilio-corchado/

More products digital (pdf, epub, mobi) instant download maybe you interests ...

Intelligent Data Engineering and Automated Learning

IDEAL 2018 19th International Conference Madrid Spain

November 21 23 2018 Proceedings Part I Hujun Yin

https://textbookfull.com/product/intelligent-data-engineeringand-automated-learning-ideal-2018-19th-international-conferencemadrid-spain-november-21-23-2018-proceedings-part-i-hujun-yin/

Intelligent Data Engineering and Automated Learning

IDEAL 2018 19th International Conference Madrid Spain

November 21 23 2018 Proceedings Part II Hujun Yin

https://textbookfull.com/product/intelligent-data-engineeringand-automated-learning-ideal-2018-19th-international-conferencemadrid-spain-november-21-23-2018-proceedings-part-ii-hujun-yin/

Product Focused Software Process Improvement 15th

International Conference PROFES 2014 Helsinki Finland

December 10 12 2014 Proceedings 1st Edition Andreas Jedlitschka

https://textbookfull.com/product/product-focused-softwareprocess-improvement-15th-international-conferenceprofes-2014-helsinki-finland-december-10-12-2014-proceedings-1stedition-andreas-jedlitschka/

Computer Safety Reliability and Security 33rd

International Conference SAFECOMP 2014 Florence Italy

September 10 12 2014 Proceedings 1st Edition Andrea Bondavalli

https://textbookfull.com/product/computer-safety-reliability-andsecurity-33rd-international-conference-safecomp-2014-florenceitaly-september-10-12-2014-proceedings-1st-edition-andreabondavalli/

Adaptive and Intelligent Systems Third International Conference ICAIS 2014 Bournemouth UK September 8 10 2014 Proceedings 1st Edition Abdelhamid Bouchachia (Eds.)

https://textbookfull.com/product/adaptive-and-intelligentsystems-third-international-conference-icais-2014-bournemouth-ukseptember-8-10-2014-proceedings-1st-edition-abdelhamidbouchachia-eds/

Hybrid Artificial Intelligence Systems 9th

International Conference HAIS 2014 Salamanca Spain June 11 13 2014 Proceedings 1st Edition Marios Polycarpou

https://textbookfull.com/product/hybrid-artificial-intelligencesystems-9th-international-conference-hais-2014-salamanca-spainjune-11-13-2014-proceedings-1st-edition-marios-polycarpou/

Intelligent Data Engineering and Automated Learning

IDEAL 2019 20th International Conference Manchester UK November 14 16 2019 Proceedings Part I Hujun Yin

https://textbookfull.com/product/intelligent-data-engineeringand-automated-learning-ideal-2019-20th-international-conferencemanchester-uk-november-14-16-2019-proceedings-part-i-hujun-yin/

Intelligent Data Engineering and Automated Learning

IDEAL 2019 20th International Conference Manchester UK November 14 16 2019 Proceedings Part II Hujun Yin

https://textbookfull.com/product/intelligent-data-engineeringand-automated-learning-ideal-2019-20th-international-conferencemanchester-uk-november-14-16-2019-proceedings-part-ii-hujun-yin/

Highlights of Practical Applications of Heterogeneous

Multi Agent Systems The PAAMS Collection PAAMS 2014 International Workshops Salamanca Spain June 4 6 2014 Proceedings 1st Edition Juan M. Corchado

https://textbookfull.com/product/highlights-of-practicalapplications-of-heterogeneous-multi-agent-systems-the-paamscollection-paams-2014-international-workshops-salamanca-spainjune-4-6-2014-proceedings-1st-edition-juan-m-corchado/

Emilio Corchado

José

Héctor Quintián

Hujun Yin (Eds.)

Intelligent Data Engineering and Automated Learning –IDEAL 2014

15th International Conference Salamanca, Spain, September 10–12, 2014 Proceedings

LectureNotesinComputerScience8669

CommencedPublicationin1973

FoundingandFormerSeriesEditors: GerhardGoos,JurisHartmanis,andJanvanLeeuwen

EditorialBoard

DavidHutchison LancasterUniversity,UK

TakeoKanade CarnegieMellonUniversity,Pittsburgh,PA,USA

JosefKittler UniversityofSurrey,Guildford,UK

JonM.Kleinberg CornellUniversity,Ithaca,NY,USA

AlfredKobsa UniversityofCalifornia,Irvine,CA,USA

FriedemannMattern ETHZurich,Switzerland

JohnC.Mitchell StanfordUniversity,CA,USA

MoniNaor

WeizmannInstituteofScience,Rehovot,Israel

OscarNierstrasz UniversityofBern,Switzerland

C.PanduRangan IndianInstituteofTechnology,Madras,India

BernhardSteffen TUDortmundUniversity,Germany

DemetriTerzopoulos UniversityofCalifornia,LosAngeles,CA,USA

DougTygar UniversityofCalifornia,Berkeley,CA,USA

GerhardWeikum MaxPlanckInstituteforInformatics,Saarbruecken,Germany

Intelligent DataEngineeringand AutomatedLearning–IDEAL2014

15thInternationalConference

Salamanca,Spain,September10-12,2014

Proceedings

VolumeEditors

EmilioCorchado

HéctorQuintián

UniversityofSalamanca PlazadelaMercedS/N 37008Salamanca,Spain,

E-mail:{escorchado,hector.quintian}@usal.es

JoséA.Lozano

UniversityoftheBasqueCountry PaseoManueldeLardizábal1 20018SanSebastián,Spain

E-mail:ja.lozano@ehu.es

HujunYin

UniversityofManchester SackvilleStreet Manchester,M139PL,UK

E-mail:hujun.yin@manchester.ac.uk

ISSN0302-9743e-ISSN1611-3349 ISBN978-3-319-10839-1e-ISBN978-3-319-10840-7 DOI10.1007/978-3-319-10840-7

SpringerChamHeidelbergNewYorkDordrechtLondon

LibraryofCongressControlNumber:2014946729

LNCSSublibrary:SL3–InformationSystemsandApplication,incl.Internet/Web andHCI

©SpringerInternationalPublishingSwitzerland2014

Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpartof thematerialisconcerned,specificallytherightsoftranslation,reprinting,reuseofillustrations,recitation, broadcasting,reproductiononmicrofilmsorinanyotherphysicalway,andtransmissionorinformation storageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilarmethodology nowknownorhereafterdeveloped.Exemptedfromthislegalreservationarebriefexcerptsinconnection withreviewsorscholarlyanalysisormaterialsuppliedspecificallyforthepurposeofbeingenteredand executedonacomputersystem,forexclusiveusebythepurchaserofthework.Duplicationofthispublication orpartsthereofispermittedonlyundertheprovisionsoftheCopyrightLawofthePublisher’slocation, inistcurrentversion,andpermissionforusemustalwaysbeobtainedfromSpringer.Permissionsforuse maybeobtainedthroughRightsLinkattheCopyrightClearanceCenter.Violationsareliabletoprosecution undertherespectiveCopyrightLaw.

Theuseofgeneraldescriptivenames,registerednames,trademarks,servicemarks,etc.inthispublication doesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfromtherelevant protectivelawsandregulationsandthereforefreeforgeneraluse.

Whiletheadviceandinformationinthisbookarebelievedtobetrueandaccurateatthedateofpublication, neithertheauthorsnortheeditorsnorthepublishercanacceptanylegalresponsibilityforanyerrorsor omissionsthatmaybemade.Thepublishermakesnowarranty,expressorimplied,withrespecttothe materialcontainedherein.

Typesetting: Camera-readybyauthor,dataconversionbyScientificPublishingServices,Chennai,India

Printedonacid-freepaper

SpringerispartofSpringerScience+BusinessMedia(www.springer.com)

Preface

TheIDEALconferenceattr actsinternationalexperts,researchers,leading academics,practitioners,andindustrialistsfromcommunitiesofmachinelearning,computationalintelligence,datamining,knowledgemanagement,biology, neuroscience,bio-inspiredsystemsand agents,anddistributedsystems.Ithas enjoyedavibrantandsuccessfulhistoryinthelast16years,havingbeenheld inover13locationsinsevendifferentcountries.

Itcontinuestoevolvetoembraceemergingtopicsandexcitingtrends.This yearIDEALwasheldinthehistoricalcityofSalamanca(Spain),declareda UNESCOWorldHeritageSitein1988andEuropeanCapitalofCulturein2002.

Thisconferencereceivedabout120submissions,whichwererigorouslypeerreviewedbytheProgramCommitteemembers.Onlythepapersjudgedtobeof highestqualitywereacceptedand includedintheseproceedings.

Thisvolumecontains60papersacceptedandpresentedatthe15thInternationalConferenceonIntelligentDataEngineeringandAutomatedLearning (IDEAL2014),heldduringSeptember10–12,2014,inSalamanca,Spain.These papersprovidedavaluablecollectionofrecentresearchoutcomesindataengineeringandautomatedlearning,frommethodologies,frameworks,andtechniquestoapplicationsandcasestudies.Thetechniquesincludecomputational intelligence,bigdataanalytics,socialmediatechniques,multi-objectiveoptimization,regression,classification,clustering,biologicaldataprocessing,text processing,andimage/videoanalysis.

Theprocessofreviewingandselectingpapersisextremelyimportanttomaintainingthehighqualityoftheconferenceandthereforewewouldliketothank theProgramCommitteefortheirhardworkintherigorousreviewingprocess. TheIDEALconferencewouldnotexistwithouttheirhelpandprofessionalism.

IDEAL2014enjoyedoutstandingkeynotespeechesbydistinguishedguest speakers:Prof.FranciscoHerreraofUniversityofGranada(Spain),Prof.Cesare AlippiofPolitecnicodiMilano(Italy),Prof.JuanManuelCorchadoofUniversity ofSalamanca(Spain),andProf.JunWangoftheChineseUniversityofHong Kong(HongKong).

Ourparticularthanksalsogotothe conferencemainsponsors,theIEEE SpainSection,theIEEESystems,ManandCyberneticsSocietySpanishChapter,AEPIA,SalamancaCityHall,CiudadRodrigoCityHall,UniversityofSalamanca,andRACEproject(USAL-USP),whojointlycontributedinanactive

andconstructivemannertothesuccessofthisconference.Finally,wewould liketothankAlfredHofmannandAnnaKramerofSpringerfortheircontinued collaborationonthispublication.

September2014EmilioCorchado Jos´eA.Lozano H´ectorQuinti´an HujunYin

Organization

HonoraryChairs

AlfonsoFern´andezMa˜nuecoMayorofSalamanca,Spain

DanielHern´andezRuip´erezRectorofUniversityofSalamanca,Spain

MariosPolycarpouUniversityofCyprus,Cyprus

GeneralChair

HujunYinUniversityofManchester,UK

ProgramChair

EmilioCorchadoUniversityofSalamanca,Spain

ProgramCo-chair

JoseA.LozanoUniversityoftheBasqueCountry,Spain

PublicityCo-chairs

JoseA.CostaUniversidadeFederaldoRioGrandedoNorte, Brazil

YangGaoNanjingUniversity,China

MinhoLeeKyungpookNationalUniversity,Korea

BinLiUniversityofScienceandTechnologyofChina, China

InternationalAdvisoryCommittee

LeiXuChineseUniversityofHongKong,HongKong, SARChina

YaserAbu-MostafaCALTECH,USA

Shun-ichiAmariRIKEN,Japan

MichaelDempsterUniversityofCambridge,UK

NickJenningsUniversityofSouthampton,UK

Soo-YoungLeeKAIST,SouthKorea

ErkkiOjaHelsinkiUniversityofTechnology,Finland

LatitM.PatnaikIndianInstituteofScience,India BurkhardRostColumbiaUniversity,USA XinYaoUniversityofBirmingham,UK

SteeringCommittee

HujunYin(Chair)UniversityofManchester,UK

EmilioCorchado(Co-chair)UniversityofSalamanca,Spain LaiwanChanChineseUniversityofHongKong,HongKong, SARChina

GuilhermeBarretoUniversityofBrazil Yiu-mingCheungHongKongBaptistUniversity,HongKong, SARChina

JoseA.CostaFederalUniversityNatal,Brazil ColinFyfeUniversityof TheWestofScotland,UK

MarcvanHulleK.U.Leuven,Belgium

SamuelKaskiHelsinkiUniversityofTechnology,Finland JohnKeaneUniversityofManchester,UK JimmyLeeChineseUniversityofHongKong,HongKong, SARChina

MalikMagdon-IsmailRensselaerPolytechnicInstitute,USA VicRayward-SmithUniversityofEastAnglia,UK PeterTinoUniversityofBirmingham,UK ZhengRongYangUniversityofExeter,UK NingZhongMaebashiInstituteofTechnology,Japan

ProgramCommittee

AboulEllaHassanienCairoUniversity,Egypt Adri˜aoDuarteFederalUniversity,UFRN,Brazil AgnaldoJos´edaR.ReisFederalUniversity,UFOP,Brazil AjalmarRˆegodaRochaNetoFederalUniversity,UFC,Brazil AjithAbrahamMirLabs,USA AlbertoGuillenUniversityofGranada,Spain AlfredoCuzzocreaUniversityofCalabria,Italy AlfredoVellidoUniversityPolit´ecnicadeCatalu˜na,Spain AliciaTroncosoUniversidadPablodeOlavide,Spain ´ AlvaroHerreroUniversityofBurgos,Spain AnaBel´enGilUniversityofSalamanca,Spain AndreCarvalhoUniversityofS˜aoPaulo,Brazil Andr´eCoelhoUniversityofFortaleza,Brazil AndreasK¨onigUniversityofKaiserslautern,Germany AndrzejCichockiBrainScienceInstitute,Japan

AnilNerodeCornellUniversity,USA AnneCanutoFederalUniversity,UFRN,Brazil AnneH˚akanssonUppsalaUniversity,Sweden

AntˆoniodeP.BragaFederalUniversity,UFMG,Brazil

AntonioNemeUniversidadAutonomadelaCiudaddeMexico, Mexico

AtaKabanUniversityofBirmingham,UK

BarbaraHammerUniversityofBielefeld,Germany BernarddeBaetsGhentUniversity,Belgium

BernardeteRibeiroUniversityofCoimbra,Portugal BinLiUniversityofScienceandTechnologyofChina, China

BogdanGabrysBournemouthUniversity,UK BrunoApolloniUniversityofMilan,Italy

BrunoBaruqueUniversityofBurgos,Spain

CarlaM¨oller-LevetUniversityofManchester,UK CarlosPereiraISEC,Portugal

CarmeloJ.A.BastosFilhoUniversityofPernambuco,POLI,Brazil

Chung-MingOuKainanUniversity,Taiwan

ChunlinChenNanjingUniversity,China

ClodoaldoA.M.LimaUniversityofS˜aoPaulo,Brazil

DanDumitrescuUniversityofBabes-Bolyai,Romania

DanielA.KeimUniversit¨atKonstanz,Germany

DanielGlez-Pe˜naUniversityofVigo,Spain DanteI.TapiaUniversityofSalamanca,Spain

DarrylCharlesUniversityofUlster,UK

DavidCamachoUniversidadAut´onomadeMadrid,Spain DavidHoyleUniversityofManchester,UK DavideAnguitaUniversityofGenoa,Italy DongqingWeiShanghaiJiaotongUniversity,China DuZhangCaliforniaStateUniversity,USA

EijiUchinoYamaguchiUniversity,Japan EmilioM.HernandezUniversityofS˜aoPaulo,Brazil

ErnestoCuadros-VargasUniversidadCat´olicaSanPablo,Peru ErnestoDamianiUniversityofMilan,Italy

EstevamHruschkaJuniorUFSCar-FederalUniversityofSaoCarlos, Brazil

EvaLorenzoUniversityofVigo,Spain FabriceRossiNationalInstituteofResearchonComputer ScienceandAutomatic,France

FelipeM.G.Fran¸caFederalUniversity,UFRJ,Brazil FernandoBuarqueUniversityofPernambuco,POLI,Brazil FernandoD´ıazUniversityofValladolid,Spain FernandoGomideUnicamp,Brazil FlorentinoFdez-Riverola UniversityofVigo,Spain FrancescoCoronaAaltoUniversity,Finland FranciscoAssisFederalUniversity,UFCG,Brazil FranciscoFerrerUniversityofSeville,Spain

FranciscoHerreraUniversityofGranada,Spain FrankKlawonnOstfaliaUniversityofAppliedSciences, Germany

GaryFogelNaturalSelection,USA GavinBrownUniversityofManchester,UK

G´erardDreyfus

´

EcoleSup´erieuredePhysiqueetdeChimie IndustriellesdeParis,France

GiancarloMauriUniversityofMilano-Bicocca,Italy

H´ectorQuinti´anUniversityofSalamanca,Spain HeloisaCamargoFederalUniversity,UFSCar,Brazil

HonghaiLiuUniversityofPortsmouth,UK

HuiyuZhouQueen’sUniversityBelfast,UK HyoseopShinKonkukUniversitySeoul,Korea IgnacioRojasUniversityofGranada,Spain

IgorFarkasComeniusUniversityinBratislava,Slovakia I˜nakiInzaUniversityofPaisVasco,Spain

IoannisHatzilygeroudisUniversityofPatras,Greece IvanSilvaFederalUniversity,USP,Brazil

J.MichaelHerrmannUniversityofEdinburgh,UK

JaakkoHollm´enHelsinkiUniversityofTechnology,Finland JaimeCardosoUniversityofPorto,Portugal JamesHoganQueenslandUniversityofTechnology,Australia

JavierBajoP´erezUniversidadPolit´ecnicadeMadrid,Spain

JavierSedanoInstitutoTecnol´ogicodeCastillayLe´on,Spain

JerzyGrzymala-BusseUniversityofKansas,USA JesusAlcala-FdezUniversityofGranada,Spain

JingLiuXidianUniversity,China

JoaoE.KoglerJr.UniversityofS˜aoPaulo,Brazil JochenEinbeckDurhamuniversity,UK JohnGanUniversityofEssex,UK

JohnQiangUniversityofEssex,UK

JonganParkChosunUniversity,Korea JorgePosadaVICOMTech,Spain

JoseA.LozanoUniversityoftheBasqueCountry,UPV/EHU, Spain

JoseAlfredoF.CostaFederalUniversity,UFRN,Brazil

Jos´eC.PrincipeUniversityofFlorida,USA Jos´eC.RiquelmeUniversityofSeville,Spain

JoseDorronsoroAut´onomadeMadridUniversity,Spain

Jos´eEverardoB.MaiaStateUniversityofCear´a,Brazil

Jos´eF.Mart´ınezInstitutoNacionaldeAstrofisicaOpticay Electronica,Mexico

Jos´eLuisCalvoRolleUniversityofACoru˜na,Spain

JoseM.MolinaUniversidadCarlosIIIdeMadrid,Spain

Jos´eManuelBen´ıtezUniversityofGranada,Spain

Jos´eRam´onVillarUniversityofOviedo,Spain

Jos´eRiquelmeUniversityofSeville,Spain

JoseSantosUniversityofACoru˜na,Spain

JuanBot´ıaUniversityofMurcia,Spain

JuanJ.FloresUniversidadMichoacanadeSanNicolasde Hidalgo,Mexico

JuanManuelG´orrizUniversityofGranada,Spain

Ju´anPav´onUniversidadComplutensedeMadrid,Spain

JuhaKarhunenAaltoUniversitySchoolofScience,Finland KeTangUniversityofScienceandTechnologyofChina, China

KeshavDahalUniversityofBradford,UK

KunihikoFukushimaKansaiUniversity,Japan LakhmiJainUniversityofSouthAustralia,Australia

LarsGraeningHondaResearchInstituteEurope,Germany

LeandroAugustodaSilvaMackenzieUniversity,Brazil

LeandroCoelhoPUCPR/UFPR,Brazil

LenkaLhotskaCzechTechnicalUniversity,CzechRepublic

LipoWangNanyangTechnologicalUniversity,Singapore LourdesBorrajoUniversityofVigo,Spain

Luc´ıaIsabelPassoniUniversidadNacionaldeMardelPlata, Argentina

LuisAlonsoUniversityofSalamanca,Spain

LuizPereiraCalˆobaFederalUniversity,UFRJ,Brazil LuonanChenShanghaiUniversity,China MaciejGrzendaWarsawUniversityofTechnology,Poland ManuelGra˜naUniversityofPaisVasco,Spain

MarceloA.CostaUniversidadeFederaldeMinasGerais,Brazil MarcinGorawskiSilesianUniversityofTechnology,Poland M´arcioLeandroGon¸calvesPUC-MG,Brazil

MarcusGallagherTheUniversityofQueensland,Australia

MariaJoseDelJesusUniversidaddeJa´en,Spain

MarioKoeppenKyushuInstituteofTechnology,Japan MariosM.PolycarpouUniversityofCyprus,Cyprus

MarkGirolamiUniversityofGlasgow,UK MarleyVellascoPontificalCatholicUniversityofRiodeJaneiro, Brazil

MatjazGamsJozefStefanInstituteLjubljana,Slovenia MatthewCaseyUniversityofSurrey,UK

MichaelHerrmannUniversityofEdinburgh,UK MichaelSmallTheUniversityofWesternAustralia,Australia MichalWozniakWroclawUniversityofTechnology,Poland MingYangNanjingNormalUniversity,China

MiroslavKarnyAcademyofSciencesofCzechRepublic, CzechRepublic NicolettaDess`ıUniversityofCagliari,Italy

OlliSimulaAaltoUniversity,Finland

OscarCastilloTijuanaInstituteofTechnology,Mexico

PabloEstevezUniversityofChile,Chile

PauloAdeodatoFederalUniversityofPernambucoand NeuroTechLtd.,Brazil

PauloCortezUniversityofMinho,Portugal

PauloLisboaLiverpoolJohnMooresUniversity,UK

PeiLingLaiSouthernTaiwanUniversity,Taiwan

PerfectoRegueraUniversityofLeon,Spain

PeterTinoUniversityofBirmingham,UK

PetroGopychUniversalPowerSystemsUSA-UkraineLLC, Ukraine

RafaelCorchueloUniversityofSeville,Spain

RamonRizoUniversidaddeAlicante,Spain

Ra´ulCruz-BarbosaTecnologicalUniversityoftheMixteca,Mexico

Ra´ulGir´aldezPablodeOlavideUniversity,Spain

RegivanSantiagoUFRN,Brazil

RenatoTin´osUSP,Brazil

RicardoDelOlmoUniversidaddeBurgos,Spain

RicardoLindenFSMA,Brazil

RicardoTanscheitPUC-RJ,Brazil

RichardChbeirBourgogneUniversity,France

RichardFreemanCapgemini,UK

RobertoRuizPablodeOlavideUniversity,Spain

RodolfoZuninoUniversityofGenoa,Italy RomisAttuxUnicamp,Brazil

RonYangUniversityofExeter,UK

RonaldYagerMachineIntelligenceInstitute-IonaCollege, USA

RoqueMar´ınUniversityofMurcia,Spain

RudolfKruseOtto-von-Guericke-Universit¨atMagdeburg, Germany

SalvadorGarc´ıaUniversityofJa´en,Spain

SamanHalgamugeTheUniversityofMelbourne,Australia

SarajaneM.PeresUniversityofS˜aoPaulo,Brazil

SeungjinChoiPOSTECH,Korea

SongcanChenNanjingUniversityofAeronauticsand Astronautics,China

StefanWermterUniversityofSunderland,UK

StelvioCimatoUniversityofMilan,Italy

StephanPareigisHamburgUniversityofAppliedSciences, Germany

Sung-BaeChoYonseiUniversity,Korea

Sung-HoKimKAIST,Korea

TakashiYoneyamaITA,Brazil

TianshiChenChineseAcademyofSciences,China

TimNattkemperUniversityofBielefeld,Germany Tzai-DerWangChengShiuUniversity,Taiwan

UrszulaMarkowska-KaczmarWroclawUniversityofTechnology,Poland VasantHonavarIowaStateUniversity,USA VasilePaladeCoventryUniversity,UK VicenteBottiPolytechnicUniversityofValencia,Spain

VicenteJulianUniversidadPolit´ecnicadeValencia,Spain Wei-ChiangSamuelsonHongOrientalInstituteofTechnology,Taiwan WeishanDongIBMResearch,China WenjiaWangUniversityofEastAnglia,UK WenjianLuoUniversityofScienceandTechnologyofChina, China

WuYingNorthwesternUniversity,USA YangGaoNanjingUniversity,China

YaniraDelRosarioDe

PazSantanaUniversidaddeSalamanca,Spain YingTanPekingUniversity,China YusukeNojimaOsakaPrefectureUniversity,Japan

LocalOrganizingCommittee

EmilioCorchadoUniversityofSalamanca,Spain

H´ectorQuinti´anUniversityofSalamanca,Spain ´ AlvaroHerreroUniversityofBurgos,Spain BrunoBaruqueUniversityofBurgos,Spain Jos´eLuisCalvoUniversityofCoru˜na,Spain

MLeNN:AFirstApproachtoHeuristicMultilabelUndersampling ..... 1 FranciscoCharte,AntonioJ.Rivera,Mar´ıaJ.delJesus,and FranciscoHerrera

DevelopmentofEye-BlinkControlledApplicationforPhysically HandicappedChildren ............................................

IppeiTorii,KaorukoOhtani,ShunkiTakami,andNaohiroIshii

GenerationofReductsBasedonNearestNeighborRelation

NaohiroIshii,IppeiTorii,KazunoriIwata,andToyoshiroNakashima

AutomaticContentRelatedFeedbackforMOOCsBasedonCourse DomainOntology ................................................

SafwanShatnawi,MohamedMedhatGaber,andMihaelaCocea

UserBehaviorModelinginaCellularNetworkUsingLatentDirichlet

RitwikGiri,HeesookChoi,KevinSooHoo,andBhaskarD.Rao

SampleSizeIssuesintheChoicebetweentheBestClassifierandFusion byTrainableCombiners

SarunasRaudys,GiorgioFumera,AistisRaudys,andIgnazioPillai

OnInterlinkingLinkedDataSourcesbyUsingOntologyMatching

AnaI.Torre-Bastida,EstherVillar-Rodriguez,JavierDelSer, DavidCamacho,andMartaGonzalez-Rodriguez

ManagingBorderlineandNoisyExamplesinImbalancedClassification byCombiningSMOTEwithEnsembleFiltering

Jos´eA.S´aez,Juli´anLuengo,JerzyStefanowski,and FranciscoHerrera

TweetSemMiner:AMeta-TopicIdentificationModelforTwitterUsing

H´ectorD.Men´endez,CarlosDelgado-Calle,andDavidCamacho

UseofEmpiricalModeDecompositionforClassificationofMRCP BasedTaskParameters

AliHassan,HassanAkhtar,MuhammadJunaidKhan,FarhanRiaz, FaizaHassan,ImranNiazi,MadsJochumsen,andKimDremstrup

DiversifiedRandomForestsUsingRandomSubspaces 85 KhaledFawagreh,MohamedMedhatGaber,andEyadElyan

FastFrequentPatternDetectionUsingPrimeNumbers

KonstantinosF.Xylogiannopoulos,OmarAddam, PanagiotisKarampelas,andRedaAlhajj

Multi-stepForecastBasedonModifiedNeuralGasMixture

AutoregressiveModel

YicunOuyangandHujunYin

LBPandMachineLearningforDiabeticRetinopathyDetection

JorgedelaCalleja,LourdesTecuapetla,Ma.AuxilioMedina, EverardoB´arcenas,andArgeliaB.UrbinaN´ajera

AutomaticValidationofFlowmeterDatainTransportWaterNetworks: ApplicationtotheATLLcWaterNetwork

DiegoGarcia,JosebaQuevedo,Vicen¸cPuig,JordiSaludes, SantiagoEspin,JaumeRoquet,andFernandoValero

Mar´ıaLuzAlonso,SilviaGonz´alez,Jos´eRam´onVillar, JavierSedano,Joaqu´ınTer´an,EstrellaOrdax,and Mar´ıaJes´usComa

CPSOAppliedintheOptimizationofaSpeechRecognitionSystem

AmandaAbelardo,WashingtonSilva,andGinalberSerra

TahaniAlqurashiandWenjiaWang

ANovelRecursiveKernel-BasedAlgorithmforRobustPattern

Jos´eDanielA.Santos,C´esarLincolnC.Mattos,and GuilhermeA.Barreto

DaniloAvilarSilvaandAjalmarRˆegoRochaNeto

PixelClassificationandHeuristicsforFacialFeatureLocalization ...... 167 HeitorB.Chris´ostomo,Jos´eE.B.Maia,andThelmoP.deAraujo

ANewAppearanceSignatureforRealTimePersonRe-identification 175 MayssaFrikha,EmnaFendri,andMohamedHammami

ANewHandPostureRecognizerBasedonHybridWaveletNetwork IncludingaFuzzyDecisionSupportSystem 183 TahaniBouchrika,OlfaJemai,MouradZaied,andChokriBenAmar

Sim-EA:AnEvolutionaryAlgorithmBasedonProblemSimilarity 191 KrzysztofMichalak

MultiobjectiveDynamicConstrainedEvolutionaryAlgorithmfor ControlofaMulti-segmentArticulatedManipulator .................. 199 KrzysztofMichalak,PatrykFilipiak,andPiotrLipinski

ParameterDependenceinCumulativeSelection ...................... 207 DavidH.Glass

ExplanatoryInferenceunderUncertainty ........................... 215 DavidH.GlassandMarkMcCartney

ANovelEgo-CenteredAcademic CommunityDetectionApproachvia FactorGraphModel .............................................. 223 YushengJia,YangGao,WanqiYang,JingHuo,andYinghuanShi

IntelligentPromotionsRecommendationSystemforInstaprom Platform ........................................................ 231

MarcosMart´ınPozo,Jos´eAntonioIglesias,and AgapitoIsmaelLedezma

Kernel K -MeansLowRankApproximationforSpectralClustering andDiffusionMaps ..............................................

CarlosM.Ala´ız, ´ AngelaFern´andez,YvonneGala,and Jos´eR.Dorronsoro

LinearRegressionFisherDiscriminationDictionaryLearningfor HyperspectralImageClassification .................................

239

247 LiangChen,MingYang,ChengDeng,andHujunYin

Ensemble-DistributedApproachinClassificationProblemSolutionfor IntrusionDetectionSystems ....................................... 255 VladimirBukhtoyarovandVadimZhukov

WeightUpdateSequenceinMLPNetworks ......................... 266 MiroslawKordos,AndrzejRusiecki,TomaszKami´nski,and KrzysztofGre´ n

ModelingofBicomponentMixingSystemUsedintheManufactureof WindGeneratorBlades ...........................................

EstebanJove,H´ectorAl´aiz-Moret´on,Jos´eLuisCasteleiro-Roca, EmilioCorchado,andJos´eLuisCalvo-Rolle

275

BranchingtoFindFeasibleSolutionsinUnmannedAirVehicleMission Planning 286

CristianRam´ırez-Atencia,GemaBello-Orgaz, MariaD.R-Moreno,andDavidCamacho

TowardsDataMartBuildingfromSocialNetworkforOpinion Analysis 295 ImenMoallaandAhlemNabli

ANovelSelfSuppressionOperatorUsedinTMA 303 JunganChen,ShaoZhongZhang,andYutianLiu

MachineLearningMethodsforMortalityPredictionofPolytraumatized PatientsinIntensiveCareUnits–DealingwithImbalancedand High-DimensionalData ........................................... 309 Mar´ıaN.MorenoGarc´ıa,JavierGonz´alezRobledo, F´elixMart´ınGonz´alez,FernandoS´anchezHern´andez,and MercedesS´anchezBarba

NonconvexFunctionsOptimizationUsinganEstimationofDistribution AlgorithmBasedonaMultivariateExtensionoftheClaytonCopula ... 318 HaroldD.deMelloJr.,Andr´eV.AbsdaCruz,and MarleyM.B.R.Vellasco

NewsMiningUsingEvolvingFuzzySystems ......................... 327 Jos´eAntonioIglesias,AlexandraTiemblo, AgapitoIsmaelLedezma,andAraceliSanchis

PredictingStudents’ResultsUsingRoughSetsTheory

AncaUdristoiu,StefanUdristoiu,andElviraPopescu

Graph-BasedObjectClassDiscoveryfromImageswithMultiple Objects 344 TakayukiNanbuandHisashiKoga

AProposedExtremeLearningMachinePruningBasedontheLinear CombinationoftheInputDataandtheOutputLayerWeights 354 LeonardoD.Tavares,RodneyR.Saldanha, DouglasA.G.Vieira,andAdrianoC.Lisboa

ADrowsyDriverDetectionSystemBasedonaNewMethodofHead

InesTeyeb,OlfaJemai,MouradZaied,andChokriBenAmar

ACBR-BasedGameRecommenderforRehabilitationVideogamesin

LauraCatal´a,VicenteJuli´an,andJos´e-AntonioGil-G´omez

ANewSemanticApproachforCBIRBasedonBetaWaveletNetwork ModelingShapeRefinedbyTextureandColorFeatures

AsmaElAdel,RidhaEjbali,MouradZaied,andChokriBenAmar

TacklingAntColonyOptimizationMeta-HeuristicasSearchMethod inFeatureSubsetSelectionBasedonCorrelationorConsistency

AntonioJ.Tall´on-BallesterosandJos´eC.Riquelme

EventStory:EventDetectionUsingTwitterStreamBased

SorabBishtandDurgaToshniwal

UnivariateMarginalDistributionAlgorithmwithMarkovChain PredictorinContinuousDynamicEnvironments

PatrykFilipiakandPiotrLipinski

ADiversity-AdaptiveHybridEvolutionaryAlgorithmtoSolvea ProjectSchedulingProblem

VirginiaYannibelliandAnal´ıaAmandi

ComputingPlatformsforLarge-ScaleMulti-AgentSimulations: TheNicheforHeterogeneousSystems

WorawanMarurngsith

BogdanWalek,Jiˇr´ıBartoˇs,andRadimFarana

DavidGriol,AraceliSanchisdeMiguel,andJos´eManuelMolina

MomodouL.SanyangandAtaKaban

AdrianLancucki,JanChorowski,KrzysztofMichalak, PatrykFilipiak,andPiotrLipinski

Zero–LatencyDataWarehouseSystemBasedonParallelProcessing andCacheModule

MarcinGorawski,DamianLis,andAnnaGorawska

AleksandarStjepanovicandMiloradBanjanin

Auto-adaptationofGeneticOperatorsforMulti-objectiveOptimization intheFirefighterProblem 484

KrzysztofMichalak

BusinessandGovernmentOrganizations’AdoptionofCloud Computing ......................................................

BilalCharifandAliIsmailAwad

Erratum

Ensemble-DistributedApproachinClassificationProblemSolutionfor IntrusionDetectionSystems

VladimirBukhtoyarovandVadimZhukov

AuthorIndex ..................................................

492

MLeNN:AFirstApproach toHeuristicMultilabelUndersampling

FranciscoCharte1 ,AntonioJ.Rivera2 , Mar´ıaJ.delJesus2 ,andFranciscoHerrera1

1 Dep.ofComputerScienceandArtificialIntelligence, UniversityofGranada,Granada,Spain 2 Dep.ofComputerScience,UniversityofJa´en,Ja´en,Spain {fcharte,herrera}@ugr.es,{arivera,mjjesus}@ujaen.es http://simidat.ujaen.es,http://sci2s.ugr.es

Abstract. Learningfromimbalancedmultilabeldataisachallenging taskthathasattractedconsiderableattentionlately.Someresampling algorithmsusedintraditionalclassification,suchasrandomundersamplingandrandomoversampling,havebeenalreadyadaptedinorderto workwithmultilabeldatasets.

InthispaperMLeNN(MultiLabeleditedNearestNeighbor ),aheuristicmultilabelundersamplingalgorithmbasedonthewell-knownWilson’sEditedNearestNeighborRule,isproposed.Thesamplestoberemovedareheuristicallyselected,insteadofrandomlypicked.Theability ofMLeNNtoimproveclassificationresultsisexperimentallytested,and itsperformanceagainstmultilabelrandomundersamplingisanalyzed. Aswillbeshown,MLeNNisacompetitivemultilabelundersampling alternative,abletoenhancesignificantlyclassificationresults.

Keywords: MultilabelClassification,ImbalancedLearning,Resampling, ENN.

1Introduction

Multilabelclassification(MLC)[1]hasmanyreal-worldapplications,beinga subjectwhichhasdrawnsignificantresearchattention.Thatmostmultilabel datasets(MLDs)areimbalancedissomethingtakenforgranted.Manyexistent methodsdealwiththisproblemthroughMLCalgorithmsadaptations[2],aiming toperformsomekindofadjustmentinthetrainingphasetotakeintoaccount theimbalancednatureofMLDs.Therearealsosomeproposalsrelyingondata resampling[3],generatingnewinstancesinwhichminoritylabelsappearor deletinginstancesassociatedtothemajorityones.

Untilnow,mostofthepublishedmultilabelresamplingalgorithmsarerandombased,includingrandomundersampling(RUS).Theaimofthispaperis toproposeamultilabelundersamplingmethodwhichheuristically,ratherthan randomly,selectstheinstancesforremoving.Theheuristicisfoundedonthe EditedNearestNeighbor(ENN)rule[4]andreliesontwomeasurestoassess

E.Corchadoetal.(Eds.):IDEAL2014,LNCS8669,pp.1–9,2014. c SpringerInternationalPublishingSwitzerland2014

theimbalancelevelinMLDs,aswellas onadistancemetricbetweenthesets oflabels(labelsets)appearingineachpairofinstances.

Thebehavioroftheproposedalgorithm,calledMLeNN,willbeexperimentallyprovedbyapplyingittosixMLDs,andanalyzingtheresultsproduced bythreeMLCmethods.Thesignificanceoftheseresultswillbestatistically evaluated,andthebenefitsproducedbyMLeNNwillbedemonstrated.

Thispaperisstructuredasfollows.InSection2abriefintroductiontomultilabelclassificationandimbalancedlearningisprovided.Section3describesthe proposedmethod,whilealltheexperimentationdetailscanbefoundinSection 4.Finally,Section5offersthefinalconclusions.

2Preliminaries

Themaincharacteristicofmultilabeleddata,whencomparedwithdataused intraditionalclassification,consistsofassociatingagroupofrelevantlabelsto eachinstance,insteadofonlyoneclass. Thisgroupisasubsetofthewholeset oflabelspresentintheMLD.Therefore,thegoalonanyMLCalgorithmisto predictthesubsetoflabelswhichshouldbeassociatedtotheinstancesinan MLD.AgeneralintroductiontoMLCcan befoundin[5].Additionally,arecent reviewofMLCmethodsisofferedin[1].

Imbalancedlearningisaproblemthoroughlystudiedintraditionalclassification,andmanysolutions,basedondifferentapproaches,havebeenproposedto faceit.Thisproblememergeswhentherearemanyinstancesbelongingtosome classes(majority),butonlyafewrepresentingothers(minority).Usually,classifierstendtobebiasedtothemajorityclasses,indetrimentoftheminorityones. Acomprehensivereviewoftraditionalimbalancedlearningsolutionsisprovided in[6].

MoststudiesassumethatallMLDsare imbalanced.Atriadofmeasures directedtoassesstheimbalancelevelinMLDsareproposedin[3],alongwithtwo randomresamplingalgorithms,oneforoversampling(LP-ROS)andanoherfor undersampling(LP-RUS).Twoofthemeasureswillbedetailedinthefollowing section,sincetheheuristicusedbyMLeNNrelyonthem.Severalmethodsaimed tofacethelearningfrommultilabelimbalanceddata,basedonensemblesofMLC classifiers[7]andnon-parametricprobabilisticmodels[2],arealsoavailable.

Ingeneral,theimbalancelevelinMLDsisnoticeablyhigherthanintraditional datasets.Moreover,algorithmsaimingtocopewiththisproblemhavetotake intoaccountthateachsamplebelongstomultiplelabels.Thus,proceduressuch asthecreationofnewinstancesordeletionofexistingoneswillinfluenceseveral labels,ratherthanonlyoneclass.

3HeuristicMultilabelUndersamplingwithMLeNN

Undersamplingalgorithmsusuallyperformworsethanoversamplingones[8], sincetheycausealossofinformationbyremovinginstances.Theinformation lossisevengreaterwhenundersamplingisappliedtoMLDs,aseachremoved

MLeNN:AFirstApproachtoHeuristicMultilabelUndersampling3

sampleisrepresentingnotonlyoneclassbutseverallabels.Asaresult,choosing therightinstancestodeleteisofcriticalimportance.AdaptingENNtowork withMLDsneedstoresolvetwokeypoints,howthecandidatesareselected andhowtheclassdifferencesamongthemandtheirneighborsareconsidered. MLeNNsettlesthesepointsbyfirstlylimitingthesampleswhichcanactas candidatestothoseinwhichnoneminoritylabelappears,andsecondlydefining ametricdistancetoknowwhatisthedifferencebetweenanypairoflabelsets.

UnlikeLP-RUS[3],whichhasarandombehavior,theMLeNNalgorithmtakes thesamplestoremoveusingaheuristicbasedonthetwofollowingbases:

– Noneoftheminoritylabelscanappearintheinstancetakenasreferenceto candidatefordeletion.

– Thelabelsetofthereferenceinstancehastobedifferenttothatofitsneighbors.

ThesecondconditionisbasedontheENNrule[4],andadaptedtothemultilabelscenarioasexplainedbelow.Algorithm1showstheMLeNNalgorithm pseudo-code.Themeasuresusedandimplementationdetailsarediscussedinthe followingsubsections.

3.1CandidateSelection

Inordertochoosewhichsampleswillactascandidatesforremoving,amethod toknowwhatlabelsareinminorityisneeded.Thoseinstancesinwhichany minoritylabelappearswillneverbecandidates,avoidingthatsomeofthefew samplesrepresentingaminoritylabelarelost.

Tocompletethistask,MLeNNreliesonthemeasuresproposedin[3].Let D beanMLD, Y thefullsetoflabelsinit,and Yi thelabelsetofthei-thinstance. IRLbl (Equation1)isameasurecalculatedindividuallytoassesstheimbalance levelforeachlabel.Thehigheristhe IRLbl thelargerwouldbetheimbalance, allowingtoknowwhatlabelsareinminorityormajority. MeanIR (Equation2) istheaverage IRLbl foranMLD,usefultoestimatetheglobalimbalancelevel.

MLeNNwilliteratethroughalltheMLDsamples,takingascandidatesthose whoselabelsetdoesnotcontainanylabelwith IRLbl > MeanIR.Thisway,all theinstancescontainingaminoritylabelwillbepreserved.

Algorithm1. MLeNNalgorithmpseudo-code.

Inputs: <Dataset>D , <Threshold> HT, <NumNeighbors> NN

Outputs:Preprocesseddataset

1: foreach sample in D do

2: foreach label in getLabelset(D ) do

3: if IRLbl(label ) > MeanIR then 4:Jumptonextsample Preserveinstancewithminoritylabels

5: endif

6: endfor

7: numDifferences ← 0

8: foreach neighbor in nearestNeighbors(sample, NN ) do

9: if adjustedHammingDist(sample, neighbor ) > HT then

10: numDifferences ← numDifferences +1

11: endif

12: endfor

13: if numDifferences ≥NN /2 then

14:markForRemoving(sample )

15: endif

16: endfor

17:deleteAllMarkedSamples(D )

3.2LabelsetdifferenceEvaluation

Wilson’sENNrulehasbeenextensivelyusedintraditionalclassification.The basicideabehinditisthefollowing:

Selectasample C ascandidate.

– Lookfor C s NN nearestneighbors.Usually NN =3.

– If C classdiffersfromtheclassofatleasthalfoftheirneighbors(thatis2 when NN =3),mark C forremoving.

Sincethereisexclusivelyoneclasstocomparewith,thedifferencebetween thecandidateclassandthatofanyofitsneighborsiseither0%(sameclass)or 100%(differentclasses).Therefore,thecandidatewillberemovedwhenthereis a100%differencebetweenitsclassandtheclassofhalformoreofitsneighbors.

Table1. Differencebetweenthelabelsetsoftwoinstances

Multilabelinstanceshavemultiplelabelsassociated,thusthedifferencebetweentwosampleslabelsetscouldbe100%,butalsoanyvaluebelowthatand

above0%.LetusconsiderthetwolabelsetsshowninTable1andhowtheirdifferencescouldbeevaluated.TheybelongtoanMLDwith376differentlabels, buteachinstanceisassociatedonlyto 3or4ofthem.Thisisthecaseofthe corel5kdataset.

UsingtheHammingdistance,itcouldbeconcludedthatadifferenceof3 existsbetweenthetwolabelsets.Asthetotallength(numberoflabels)is376, thiswouldgivea0 798%difference.However,ifonlytheactivelabels(those whicharerelevant)ineitherlabelsetareconsideredtheresultwouldbetotally different,sincethereareonly7activelabelsintotal.Thepercentageofdifference wouldbe42 857%,farhigherthanthepreviousone.TherearemanyMLDswith hundredsofdistinctlabels,butalownumberofactivelabelsineachsample. CalculatingdifferencesusingtheusualHammingdistancewillalwaysproduce extremelylowvalues.Thus,consideringonlyactivelabelsmakessensewhenit comestoevaluatedifferencesamonglabelsets.

MLeNNcalculatesan adjusted Hammingdistancebetweenthecandidateand theirneighborslabelsets,countingthenumberofdifferencesanddividingit bythenumberofactivelabels.Asaresult,avalueintherangeof[0,1]is obtained.Applyingaconfigurablethreshold(HT inAlgorithm1),thealgorithm determineswhichofitsneighborswillbeconsideredasdistinct.

4ExperimentationandAnalysis

Thissectiondescribestheexperimentalframeworkusedtotesttheperformance oftheproposedalgorithm.Afterwards,obtainedresultsandtheiranalysisare provided.

4.1ExperimentalFramework

Thepaperexperimentationhasbeenstructuredintwophases.Thefirstgoalis todetermineifMLeNNisabletoimproveclassificationresults.Itcomparesthe outputofclassifiersbeforeandafterpreprocessingthesamesetofdatasets.The secondphaseaimstocompareMLeNNperformanceagainstthatofmultilabel randomundersampling,preprocessingtheoriginaldatasetswithLP-RUS[3].

ThesixdatasetswhosecharacteristicsareshowninTable2wereusedto experimentallyassessMLeNN1 .AllofthemarefromtheMULANrepository [9],andhavebeenrepeatedlyusedinthe literature.Theyhave beenpartitioned followinga2x5strategy.Ascanbeinferredfromthe MeanIR values,fiveofthese datasetsaretrulyimbalanced,withvaluesrangingfrom7.20to256.40.Onthe contrary,theemotionsdatasetcouldnotbeactuallyconsideredasimbalanced. IthasbeenincludedintheexperimentationtotestthebehaviorofMLeNN whenusedwithnon-imbalancedMLDs.MLeNNwasappliedtoallofthemwith NN =3(3neighbors)and HT =0.75(75%labelsetdifferencethreshold).

1 Thispaperhasanassociatedwebsiteat http://simidat.ujaen.es/mlenn .Both datasetpartitionsandtheMLeNNprogramcanbedownloadedfromit.Thiswebsite alsooffersfulltablesofresults.

6F.Charteetal.

RegardingtheMLCalgorithmsused,abasicBinaryRelevance(BR[10]) transformationmethodandtwomoreadvancedclassifiers,CalibratedLabel Ranking(CLR[11])andRandomk-labelsets(RAkEL[12]),wereselected.The well-knownC4.5tree-basedclassificationalgorithmwasusedasunderlyingclassifierwhereneeded.

Table2. Characteristicsofthedatasetsusedinexperimentation

4.2ResultsandAnalysis

TheoutputsproducedbytheMLCalgorithms,learningfromthebaseMLDsand thoseobtainedafterpreprocessingwithMLeNNandML-RUS,havebeenevaluatedwiththreemeasures:Accuracy,Macro-FMeasureandMicro-FMeasure. Theformerisasample-basedevaluationmeasure,thusalllabelsaregivenequal weight,whereastheothertwoarelabel-based.Macro-averagingmeasurestend toemphasizetheresultsofrarelabels,whilemicro-averagingdoestheopposite.Howthesemeasuresassesspredictionperformancecanbefoundin[5].An intuitivevisualrepresentationofthoseresultsisofferedinFigure1.

First,itcanbeseenthattheundersamplingperformedbyMLeNNhasimprovedbaseresultsinmanycases.However,therearesomeexceptions.The mostremarkableisthatoftheemotionsdataset,whoseresultstendtobeworse afterMLeNNhasbeenapplied.Thisledtotheconclusionthatundersampling, whetherrandomorheuristic,shouldnotbeappliedtoMLDswhicharenottruly imbalanced.AnotherfactthatcanbevisuallyconfirmedinFig.1isthatthe performanceofMLeNNisalmostalwaysbetterthanthatofML-RUS.

Aimingtoformallyendorsetheseresults,aWilcoxonnon-parametricstatisticaltestwasusedtocompareMLeNNwithbaseresults,firstly,andwith ML-RUS,secondly.TheresultsofthesetestsareshowninTable3andTable 4.Astarattherightofavaluedenotesthatitisthebestrankingforagiven measure.Thesymbol indicatesthatthereisnostatisticaldifferencebetween thisrankingandthebestone,whereasthesymbol statesthatasignificant differenceexists.

Fromtheanalysisoftheseresults,thatMLeNNisacompetitivemultilabel undersamplingalgorithmcanbeconcluded,sinceitalwaysachievesbetterperformancethanML-RUSfromanstatisticalpointofview.Moreover,theundersamplingconductedbyMLeNNisabletoimproveclassificationresultswhen comparedwiththoseobtainedwithoutpreprocessing.MLeNNproducedastatisticallysignificantimprovementintwoofthethreemeasures,despitetheinclusion ofanMLDsuchasemotions,whichisnotimbalanced,intothestatisticalstudy.

BaseMLeNNLP−RUS

Fig.1. Eachplotshowsclassificationresultscorrespondingtoameasure/algorithm combination

Table3. Firstphase-AverageRankings

Table4. Secondphase-AverageRankings

8F.Charteetal.

5Conclusions

ThelearningfromimbalancedMLDspresentssomeseriousdifficulties.Several approacheshavebeenproposedtoovercomethem,includingrandomundersamplingalgorithms.Thiskindofmethodsdoesnotusuallyworkwell,sincethey causeasignificantlossofinformationpotentiallyusefulinthetrainingprocess. InthispaperMLeNN,anovelmultilabelheuristicundersamplingalgorithm, hasbeenpresented.Thedeletedinstancesarethoughtfullyselected,insteadofbeingrandomlychosen.Astheobtainedresultsshow,itisatechniqueabletoimproveclassificationresultswhenappliedtotrulyimbalancedMLDs.Moreover, itperformssignificantbetterthantherandomundersamplingimplementedby LP-RUS.

Acknowledgments. F.CharteissupportedbytheSpanishMinistryofEducationundertheFPUNationalProgram(Ref.AP2010-0068).Thisworkwas partiallysupportedbytheSpanishMinistryofScienceandTechnologyunder projectsTIN2011-28488andTIN2012-33856(FEDERFounds),andtheAndalusianregionalprojectsP10-TIC-06858andP11-TIC-9704.

References

1.Zhang,M.-L.,Zhou,Z.-H.:Areviewonmulti-labellearningalgorithms.IEEE Trans.Knowl.DataEng.(2013)

2.He,J.,Gu,H.,Liu,W.:Imbalancedmulti-modalmulti-labellearningforsubcellular localizationpredictionofhumanproteinswithbothsingleandmultiplesites.PloS One7(6),7155(2012)

3.Charte,F.,Rivera,A.,delJesus,M.J.,Herrera,F.:Afirstapproachtodealwith imbalanceinmulti-labeldatasets.In:Pan,J.-S.,Polycarpou,M.M.,Wo´zniak,M., deCarvalho,A.C.P.L.F.,Quinti´an,H.,Corchado,E.(eds.)HAIS2013.LNCS, vol.8073,pp.150–160.Springer,Heidelberg(2013)

4.Wilson,D.L.:Asymptoticpropertiesofnearestneighborrulesusingediteddata. IEEETrans.onSMC-2(3),408–421(1972)

5.Tsoumakas,G.,Katakis,I.,Vlahavas,I.:MiningMulti-labelData.In:Maimon, O.,Rokach,L.(eds.)DataMiningandKnowledgeDiscoveryHandbook,ch.34, pp.667–685.SpringerUS,Boston(2010)

6.Haibo,H.,Yunqian,M.:ImbalancedLearning:Foundations,Algorithms,andApplications.Wiley-IEEEPress(2013)

7.Tahir,M.A.,Kittler,J.,Bouridane,A.:Multilabelclassificationusingheterogeneousensembleofmulti-labelclassifiers.PatternRecognit.Lett.33(5),513–523 (2012)

8.Garc´ıa,V.,S´anchez,J.,Mollineda,R.:Ontheeffectivenessofpreprocessingmethodswhendealingwithdifferentlevelsofclassimbalance.Knowl.BasedSystems25(1),13–21(2012)

9.Tsoumakas,G.,Xioufis,E.S.,Vilcek,J.,Vlahavas,I.:MULAN:AJavaLibraryfor Multi-LabelLearning.J.Mach.Learn.Res.12,2411–2414(2011)

10.Godbole,S.,Sarawagi,S.:DiscriminativeMethodsforMulti-labeledClassification. In:Dai,H.,Srikant,R.,Zhang,C.(eds.)PAKDD2004.LNCS(LNAI),vol.3056, pp.22–30.Springer,Heidelberg(2004)

11.F¨urnkranz,J.,H¨ullermeier,E.,LozaMenc´ıa,E.,Brinker,K.:Multilabelclassificationviacalibratedlabelranking.Mach.Learn.73,133–153(2008)

12.Tsoumakas,G.,Vlahavas,I.:Random k-labelsets:Anensemblemethodformultilabelclassification.In:Kok,J.N.,Koronacki,J.,LopezdeMantaras,R.,Matwin, S.,Mladeniˇc,D.,Skowron,A.(eds.)ECML2007.LNCS(LNAI),vol.4701,pp. 406–417.Springer,Heidelberg(2007)

Development of Eye-Blink Controlled Application for Physically Handicapped Children

Dept. of Information Science, Aichi Institute of Technology, Aichi, Japan {mac,ishii}@aitech.ac.jp ruko2011@gmail.com mail@takamin.net

Abstract. In this paper, we describe a new application operated with eye-blink for physically handicapped children who cannot speak to communicate with others. Process of detecting blinks is performed in the following steps. 1) To detect an eye area 2) To distinguish opening and closing of eyes 3) To add the method using saturation to detect blink 4) To decide by a conscious blink 5) To improve the accuracy of detection of a blink We reduce the error to detect a blink and pursue the high precision of the eye chasing program. The degree of disablement is varied in children. So we develop the system to be able to be customizes depends on the situation of users. And also, we develop the method into a communication application that has the accurate and high-precision blink determination system to detect letters and put them into sound.

Keywords: VOCA (Voice Output Communication Aid), Phsically handicapped children, OpenCv, Haar-like eye detection.

1 Introduction

Special support schools in Japan need some communication assistant tools especially for physically handicapped children. In this study, physically handicapped children are defined as children with permanent disablements of their trunks and limbs because of cerebral palsy, muscular dystrophy, spina bifida and so on. Their body movements are very limited and many of them also have mental disorders, so they cannot communicate with their families or caregivers. It prevents helpers from understanding what they really need or think.

Taking the situation into consideration, we develop a communication support tool operated by eye-blink for physically handicapped children. We use front cameras on tablets (iPad, iPad mini of Apple Inc. iOS5.0). After several verifications and examination, we have developed a contactless communication assistant application “Eye Talk”. The application does not malfunction by surroundings, such as brightness or differences of eye shapes.

* Corresponding author.

E. Corchado et al. (Eds.): IDEAL 2014, LNCS 8669, pp. 10–17, 2014. © Springer International Publishing Switzerland 2014

Another random document with no related content on Scribd:

Another thing alike among these early peoples, is that all of them had drums and rattles of some kind and a roughly made instrument that resembles our pipes. But they had no stringed instruments and for their beginnings you will have to journey on with us in this,—your book.

Since giving this book to the public, we have come in direct contact with some remarkable songs of the Nootka (Canadian) Indians and of the Eskimos. Juliette Gaultier de la Verendry, a young French Canadian, has sung them in New York in the original dialect. They have been given to her by D. Jenness, an anthropologist who lived among the Eskimos for several years, studying their traits and at the same time he took the opportunity of writing down their songs. They are truly savage music and have the characteristics of which we have spoken in the use of intervals, drums, and in the type of songs, such as weather and healing incantations (medicine songs), work songs, and dances.

CHAPTER III

The Ancient Nations Made Their Music—Egyptian, Assyrian, and Hebrew

Three thousand years before Jesus was born, a corner of southwestern Asia and northeastern Africa was the home of people who had reached a very high degree of civilization. They were the first to pass the stage of primitive man, and to make for themselves beautiful buildings, beautiful cities, monuments, decorations and music. Among these ancient, civilized people were the Egyptians, the Assyrians and the Hebrews. We will talk first about the Egyptians because they had the greatest influence not only on the Assyrian and Hebrew music, but also on the Greeks who went to Egypt. So, in European music we can trace the Egyptian influence through the Greeks.

The Egyptians were very fond of building and they decorated what they built with pictures in vivid colorings called hieroglyphics (heiro —sacred, glyphics—writings). As they had neither newspapers nor radio sets, they carved or painted the records of their daily lives, their festivals, battles, entertainments, and even marketing journeys on the walls and on the columns of the temples, on the obelisks, and in the tombs, some of which were the pyramids.

The climate saved these records from destruction, and the archaeologists re-discovered them for us in the tombs full of Egyptian treasure and the temples and lost cities, buried for thousands of years.

Fig. 1. Drums and Sticks.
Fig. 2. Sioux Drum.

Fig. 3.

A Pipe and Rattles from Alaska.

(Courtesy of the Metropolitan Museum of Art, New York.)

Fig. 4. Bone Flutes.

Some Instruments of the American Indian.

From the Egyptian Collection in the Louvre, Paris.

Hieroglyphics on an Egyptian Tablet. (Telling a story of a Prince.)

The Egyptians built to inspire feelings of awe, mystery and grandeur. You probably remember pictures of obelisks, temples, pyramids, tombs and sphinxes, alongside of which a man looks but a few inches high.

They were very young as the world goes, and built huge structures because they were still filled with wonder at the immensity and power of the things they saw in Nature,—the Nile; the great desert

which seemed vaster to them because they had only slow-moving camels, elephants and horses to take them about; they saw very long rainy seasons and the Nile overflowing its banks yearly, long dry seasons and the terrible wind and sand storms; the great heat of the sun, and the glory of their huge flowers, such as the lotus.

Just as primitive people did, they personified Nature in the gods. They had Osiris—god of Light, Health and Agriculture; Isis—goddess of the Arts and Agriculture; Horus (hawk-headed), the Sun god; Phtah, first divine King of Memphis, and many others. Again like primitive people, they had music for their gods, for their temple services, for their state ceremonies, festivals, martial celebrations and amusements.

Primitive music, we saw, had no laws to bind it, but was guided by the savage’s natural feeling and he could make up anything he wished. In Egypt, because of state law which prevented it from changing, music was held down to the same system for three thousand years. New music was forbidden, and much of the old was considered sacred and so closely connected with religious ceremonies that it was allowed to be used only in the temples.

The priests lived in these magnificent temples and were the philosophers, artists and musicians, very like the medicine men of the Indians, but much more advanced in learning.

Like the American Indians, too, the profession of music was handed down from father to son, and only the children of singers, whether they had good voices or not, could sing in the temples.

On the monuments we see these singers followed by players of instruments. The singers were of the highest caste, or Priest caste; the players were usually of the lower classes, or the Slave caste, although as pictured on the tombs of Rameses, one of Egypt’s greatest rulers and builders, we see the priests dressed in splendid robes and playing large harps.

The temples of Egypt were so huge that the music had to be on a large scale. They thought nothing of an orchestra of six hundred players of harps, lyres, lutes, flutes and sistrums (bell rattles), whereas we today advertise in large type the fact of one hundred men in one orchestra! We see no trumpets in the picture writings of the Egyptian orchestra, for these were only used in war, and we find

them only in their pictures of war and triumphal marches; nor do we see large drums, because the Egyptians clapped their hands to mark rhythm. However, the military instruments in the hands of players pictured on the monuments, show that they used trumpets and tambourines in the army.

From the names we find in the tombs—“Singers of the King” and “Singers of the Master of the World,” we know that the Kings had musicians of high rank in their courts. The paintings on the walls and columns of the ruins of the temple Karnak, show funeral services with kneeling singers, playing harps of seven strings and other instruments.

Ptolemy Soter II, another famous Egyptian ruler, gave a fête in which were heard a chorus of twelve hundred voices, accompanied by three hundred Greek kitharas and many flutes.

It seems like a fairy tale that we can bring back the manners and customs of three thousand years ago through studying the writings in stone called hieroglyphics, and by examining the things used every day, that were found in the excavations. For a long time the hieroglyphics were unsolved riddles until the discovery in 1799 A.D. of the Rosetta stone, on which was an inscription in hieroglyphics with its Greek translation. Although ancient Greek is called a dead language, it still has enough life in it to bring back the history and records of antiquity. Through this knowledge of Greek, the Egyptian inscriptions speak to us and tell us marvelous stories of ancient Egypt.

In one of the tombs at Thebes, was a harp with strings of catgut, which when plucked, still gave out sounds although the harp had probably not been played upon in three thousand years!

Going once more to our ancient stone library—or collections of monuments in our museums or in Egypt—we see many pictures of dancers. The Egyptians danced in religious ceremonies as well as in private entertainments. They loved lively dances, and the men did all sorts of acrobatic steps and even toe-dancing like our Pavlowa, while the women did the slow, languorous dances.

Egyptian music was greatest as far back as 3000 B.C.! After that it grew poorer until 525 B.C. when Egypt was conquered by Persia.

The Egyptians must have used a musical scale of whole steps and half steps, covering several octaves, not unlike ours. Think of the piano keyboard with its black and its white keys and you will get an idea of the Egyptian scale. We learned this through the discovery of a flute that played a scale of half steps from a below middle c to d above the staff with only a few tones missing.

A M

In the British Museum in London and in the Louvre in Paris, you can see ancient records which archaeologists unearthed from three mounds near the River Tigris in Asiatic Turkey. These mounds were the remains of the Assyrian cities of Nimroud (Babylon), Khorsabad, and probably the famous Nineveh, and date from 3000 to 1300 B.C.

Did Assyria influence Egypt or was it the other way around? The Egyptians excelled in making mechanical things such as instruments, utensils, tools, and in building temples and pyramids; while the Assyrians were sculptors, workers in metals and enamel, and knew the secret of dyeing and weaving stuffs, and of making beautiful pottery. But whose music was the better, the Egyptians or the Assyrians, is impossible to say. We do know, however, that the Assyrians, as well as the Egyptians and Hebrews, had perfected music far beyond the standard reached by many nations of our own time.

The Assyrians had the same families of instruments that we have, —the percussion (or drums), wind, and strings; and they used different combinations of instruments in concerts, either in instrumental performances or for accompanying vocal music. Everything that we know about them shows that the Assyrians were greater noisemakers than the Egyptians, for they not only had drums and trumpets, but they also marked rhythm by stamping their feet instead of clapping their hands.

The instruments pictured on the monuments, probably existed many centuries before the building of these monuments, which would make them very old indeed. In fact, almost all of them are still in use in the Orient today and are played in the same way. The monuments also prove that some of the special ceremonies in which music was used are still in existence.

Both the Assyrians and the Egyptians had flutes, and double flutes which were actually two flutes connected by one mouthpiece and looked like the letter v. The Assyrians also had harps that varied in size from some that could be carried in the hand, to some that stood seven feet high and had as many as twenty-two strings. The dulcimer, an instrument something like a zither, was very popular and was made so that it could be played standing upright or lying

flat. They also had drums, castanets, cymbals, tambours or tambourines, and lyres, all of which could be easily carried.

The Assyrians being a warlike nation made their instruments so that they could be strapped to their bodies. So it seems that people in 3000 B.C. were practical.

The Assyrians were so fond of music that when their war-prisoners were musicians they were not put to death.

H M

We get our knowledge of the Hebrew music not from stone monuments and wall pictures, but from Biblical writings and other ancient Hebrew records. In the Second Commandment, God forbids the Hebrews to make images:

“Thou shalt not make unto thyself any graven image, nor the likeness of anything that is in heaven above, or in the earth beneath, or in the waters under the earth.” (Exodus XXI: 4.) With so strict a commandment, you can understand why there are no pictures of singers and of instruments, and that we have to go to the greatest literary gift to the world,—the Old Testament, to find out about their music.

The first musician mentioned in the Bible is Jubal. It says in Genesis IV: 21, “he was the father of all such as handle the harp and pipe (or organ).” From an old Spanish book found in the early 18th century in a Mexican monastery, comes the story that Jubal was listening to Tubal-Cain’s forge, and noticed the difference in pitch of the sounds made by the strokes on the anvil. Some tones were high, some low, and some were medium. He compared this to the human voice, and tried to imitate the sounds, high, low and medium, of the forge. Thus he became the first singer of the Hebrews. Jubal invented a flute and a little three-cornered harp called the kinnor. These small instruments were most convenient to carry about, for at this time the Hebrews were shepherd tribes wandering from place to place. Their music was simple as is the music of all primitive peoples.

We know from the Biblical story that the Children of Israel were sold into captivity and remained many centuries in Egypt; that Moses was found in the bulrushes by Pharoah’s daughter, and was educated as an Egyptian boy and “was learned in all the wisdom of the Egyptians.” Therefore, he must have learned music from the priests. It is natural then, that the Hebrews must have borrowed the music and instruments of their adopted country in the making of their own.

After Moses had been commanded by the Lord to lead the children of Israel out of the land of captivity, and after the Red Sea had

divided to allow them to pass through, we read the great song of triumph sung by Moses:

“Then sang Moses and the Children of Israel: ‘I will sing unto Jehovah, for he hath triumphed gloriously, the horse and his rider hath he thrown into the sea, Jehovah is my strength and my song and he is become my salvation.’” etc.—(Exodus XV: 1–2).

And “Miriam, the prophetess, the sister of Aaron, took a timbrel in her hand, and all the women went out after her with timbrels and with dances.” (The timbrel is a small tambourine-like instrument.)

This story, like others in the Old Testament, is full of the accounts of musical instruments, singing and dancing, and shows us that the ancient Hebrews used music and the dance for nearly every event. If you read carefully you will get the musical history of this poetic people.

While the Children of Israel were in the wilderness, Moses received from Jehovah the command: (Numbers X.)

“Make thee two trumpets of silver; of a whole piece shalt thou make them; that thou mayest use them for the calling of the assembly, and for the journeyings of the camp.”

Then follow directions as to the meaning of the blowing of the trumpets. One trumpet alone called the princes; two trumpets called the entire tribe together; an “alarm” gave the signal for the camps to go forward, and so on. So, you see the ancient Hebrews used trumpets much as we today use the army bugle. The trumpets mentioned as one of the earliest of all instruments called the people to religious ceremonies too; it announced festivals, the declaration of a war, the crowning of a king, proclaimed the jubilee year, and gave warning of the anger of God.

One instrument has come down to our times and is still used in the Hebrew temple services. This is called the shofar and is usually a ram’s horn on which two tones may be blown. Probably, as the ram was one of the animals of sacrifice, they used its horn as a sacred instrument. This shofar is 5,000 years old, at least. It is sounded in all the synagogues of the world on the Jewish New Year and on the Day of Atonement in memory of the wanderings of the Children of Israel.

When the twelve tribes, after their wanderings in the wilderness, had settled down in Palestine, they gave music a most important place in their daily life. Samuel, the last and most respected of the judges, built a school of prophecy and music. Here it was that young David hid himself to escape the persecutions of Saul. You remember that David is called the Great Musician and he gave us many of the Psalms, the most beautiful religious verse in the world. How much it would mean to us if we knew the music David sang to these songs! In spite of the fact that the music in which they were originally sung has been lost, the Psalms have been an inspiration to all composers of religious music throughout the ages. David learned so much at Samuel’s school that he created a most beautiful musical service for the temple, which is the basis of the one used today in Jewish synagogues (temples).

The number that were instructed in the songs of the Lord was two hundred, four score and eight (288). There were in all four thousand, including assistants, students, players of instruments and the two hundred and eighty-eight professional singers.

All of these people did not perform at one time; for the ordinary services they used twelve male singers, twelve players on instruments,—nine harps, and two players of the psaltery and one of cymbals. Women were not allowed to sing in the temples but they were a part of the court and sang at funerals and at public festivities and banquets.

The great Jewish historian, Josephus, tells us that Solomon had two hundred thousand singers, forty thousand harpists, forty thousand sistrum players and two hundred thousand trumpeters. This is hard to believe, but as everything belonged to the kings in bygone days, probably this was only Solomon’s musical directory.

The psaltery was an instrument something like our zither, with thirteen strings on a flat wooden sounding board, rectangular in shape. The sistrum was a metal rattle which made a very sweet sound.

It isn’t easy to describe the instruments used thousands of years ago, for the names have become changed through the ages, and we find the same type of instruments called by different names in different countries and periods. For example, the psaltery is much the same instrument as the dulcimer, the Arab’s kanoun and the

Persian santir. We find the same psaltery in Chaucer’s “Miller’s Tale” as sautrie. By the addition of a keyboard this Biblical instrument became the spinet, which you will meet again in the 15th and 16th centuries. In the 13th century, in Italy, we find a kind of psaltery hung around the neck and called “Istroménto di porco” because it looked like the head of a pig.

Not all the songs of the Bible are religious. The Song of Solomon, a most beautiful poem of marriage, gives us a vivid picture of luxury and magnificence, as well as showing us that music was used for other than religious ceremonies.

After the death of Solomon, the music in the temple lost its splendor and again the Children of Israel were made captive. When one hundred years later Nebuchadnezzar (586 B.C.) the King of Babylon, destroyed their temple, the song of the Hebrews became sad and mournful, as you can read in the Book of Lamentations and in this beautiful song of grief, the 137th Psalm:

SORROWS OF THE EXILES IN BABYLON

By the rivers of Babylon

There we sat down, yea, we wept, When we remembered Zion.

Upon the willows in the midst thereof We hanged up our harps.

For there they that led us captive required of us songs, And our tormentors required of us mirth, saying: “Sing for us one of the songs of Zion.”

How shall we sing Jehovah’s song In a foreign land?

If I forget thee, Oh Jerusalem, Let my right hand forget her skill.

Let my tongue cleave to the roof of my mouth If I remember thee not; If I prefer not Jerusalem, Above my chief joy.

During the next few centuries the Hebrews became scattered over the world, carrying with them their reverence for God, their love of poetry and song, and their religious customs. These qualities have persisted throughout the centuries, and some of the greatest musicians in the world have been of Hebrew origin.

Although most of the old music has passed away, there is still enough of its spirit left in their temple services to give some idea of the ancient Hebrew music.

From a panel in a Museum (delle Terme) in Rome.
Greek Girl Playing a Double Flute (auloi).
From a frieze in the Boston Museum of Fine Arts.
Greek Boy Playing the Lyre.

CHAPTER IV

The Greeks Lived Their Music—The Romans Used Greek Patterns

The Greeks “dwelt with beauty” and believed it to be a part of being good, and they strove to make everything beautiful. Beauty to the Greeks was a religion. Had this not been so, we would not have the Venus de Milo, the Parthenon in Athens, the Hermes, the Winged Victory (Niké of Samothrace) and all the other Greek masterpieces which no modern sculptor or builder has surpassed.

It is interesting to see a nation 400 years before the time of Christ and even earlier, making glorious art works in stone, and writing the greatest plays the world has ever had, being more grown up than modern nations, and yet as far as we know an infant in the art of music. We have only the slightest idea of how their music sounded as they had no accurate way of writing it, and had only very primitive instruments. Although when compared to their other arts their music was not great, still it was very important to them and they used it constantly with poetry, dancing, and in the drama.

The word music was first used by the Greeks and has been carried into nearly every language; we find musique in French, Musik in German, musica in Italian, and so on.

Music, according to the Greeks, was an art which combined not only the playing of instruments, singing and dancing, but also all the arts and sciences, including mathematics and everything in the universe. It took its name from the Muses, and they believed that it led to the beautiful accord and harmony of the world.

The nine Muses were daughters of Jupiter, and each presided over some particular department of literature, art and science.

Clio: Muse of History and Epic Poetry. She is shown in statues and pictures holding a half open scroll.

Thalia: Muse of Joy and Comedy (drama) with a comic mask in one hand and a crooked staff in the other.

Erato: Muse of Lyric Poetry, inspired those who wrote of love. She plays on a nine-stringed lyre.

Euterpe: Muse of Lyric Song, patroness of music especially of flute players. She holds two flutes (auloi).

Polyhymnia: Muse of Sacred Song. She holds her forefinger to her lips or carries a scroll.

Calliope: Muse of Eloquence and Epic Poetry, holds a roll of parchment, or a trumpet.

Terpsichore: Muse of the Dance, presiding over choral, dance and song. She appears dancing with a seven-stringed lyre.

Urania: Muse of Astronomy, holds the globe and traces mathematical figures with a wand.

Melpomene: Muse of Tragedy (drama), leans on a club and holds a tragic mask.

The myths and legends of the ancient Greeks read like fairy tales, but to the Greeks they were what our Bible stories are to us. In their rich mythology we find many stories about the beginnings of music.

To Pan, the god of woods and fields, of flocks and shepherds, is given the credit of inventing the shepherd’s pipe, or Pan’s Pipes. He lived in grottoes, wandered on the mountains and in the valleys, and amused himself hunting, leading the dances of the nymphs, and playing on his pipes.

P’

P

A beautiful nymph named Syrinx was loved by Pan, but every time that he tried to tell her of his love, she became frightened and ran away, for Pan was a funny looking lover with goat’s legs, a man’s body, and long pointed ears. One day he chased her through the woods to the bank of a river; she called out in fright, and was suddenly changed by her friends the Water Nymphs, into a clump of tall reeds. When he reached out to embrace her, instead of Syrinx, he had the clump of reeds in his arms! As he sighed in disappointment, his breath passing through the reeds, produced a sad wail. Pan, hearing in it a plaintive song, broke off the reeds in unequal lengths, bound them together, and made the first musical instrument, which he called a syrinx in memory of his lost sweetheart. These pipes comforted Pan, and he played many tender melodies, and often without being seen, was known to be near by his lovely music.

Pan, although adored, was feared. At one time, Brennus, a warrior, with a company of Gauls (a tribe from ancient France), attacked the Temple of Delphi (in Greece), and was about to destroy it, when suddenly they turned and fled in fear although no one pursued them. Their terror was supposed to have been of Pan’s making, and to this day we use the word “panic” (Pan-ic) for all sudden overpowering fright.

Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.