KerelMethodsToolbox
AMATLABtoolboxforoliearsigalprocessigadmachielearig
Author:SteveVaVaerebergh
Officialweb:https://github.com/steve2358/kmbox
AboutTheKerelMethodsToolbox(KMBOX)isacollectioofMATLABprogramsthatimplemetkerel-basedalgorithms,withafocusoregressioalgorithmsadoliealgorithms.Itcabeusedforoliearsigalprocessigadmachielearig.
KMBOXicludesimplemetatiosofalgorithmssuchaskerelpricipalcompoetaalysis(KPCA),kerelcaoicalcorrelatioaalysis(KCCA)adkerelrecursiveleast-squares(KRLS).
Thegoalofthisdistributioistoprovideeasy-to-aalyzealgorithmimplemetatios,whichrevealtheiermechaicsofeachalgorithmadallowforquickmodificatios.Thefocusoftheseimplemetatiosisthereforeoreadabilityratherthaspeedormemoryusage.
ThestartigpoitofthistoolboxwasasetofprogramswritteforthePh.D.Thesis"KerelMethodsforNoliearIdetificatio,EqualizatioadSeparatioofSigals".
Templatefilesareprovidedtoecourageexteralauthorstoicludetheirowcodeitothetoolbox.
CopyrightoticeThecodehasbeedevelopedadcopyrighted©2014bySteveVaVaerebergh.ItisdistributeduderthetermsoftheBSD(3-Clause)Licese.Ishort,thismeasthateveryoeisfreetouseit,tomodifyitadtoredistributeitoafreebasis.Itisotithepublicdomai;itiscopyrightedadtherearerestrictiosoitsdistributio(seeLICENSE.txt).
IstallatioRuistall.mtoaddthelibraryfoldertothepath.Typesavepathtosavethechagestothepath.UsageTheameofeachfuctiousestheprefixkm_tomiimizeiterferecewithothertoolboxes.Usageofeachfuctioisspecifiedithefuctiofileitself.
Mostalgorithmshaveacorrespodigdemostratiofileithe"demo"folderthatstartswith"km_demo".Thesearescriptsthatcabeexecutedwithoutsettigayadditioalparameters.
Thecodeusesthefollowigcovetios:
Fordatamatrices,dataisstoredadaccessedirowformat:eachdatapoitisarowithedatamatrix.CitigKMBOXIfyouusethistoolboxiyourresearchpleasecitethisPh.D.thesis:
@phdthesis{vavaerebergh2010kerel,author={VaVaerebergh,Steve}title={Kerelmethodsforoliearidetificatio,equalizatioadseparatioofsigals},year={2010},school={UiversityofCatabria},moth=feb,ote={Softwareavailableat\url{https://github.com/steve2358/kmbox}}}IcludedalgorithmsKerelRidgeRegressio(KRR).PricipalCompoetAalysis(PCA).KerelPricipalCompoetAalysis(KPCA),asproposediB.Scholkopf,A.SmolaadK.R.Muller,"Noliearcompoetaalysisasakereleigevalueproblem",NeuralComputatio,volume10,o.5,pages1299-1319,1998.ApproximateLiearDepedecyKerelRecursiveLeast-Squares(ALD-KRLS),asproposediY.Egel,S.Maor,adR.Meir."Thekerelrecursiveleast-squaresalgorithm",IEEETrasactiosoSigalProcessig,volume52,o.8,pages2275–2285,2004.Slidig-WidowKerelRecursiveLeast-Squares(SW-KRLS),asproposediS.VaVaerebergh,J.Via,adI.Satamaria."Aslidig-widowkerelRLSalgorithmaditsapplicatiotooliearchaelidetificatio",2006IEEEIteratioalCofereceoAcoustics,Speech,adSigalProcessig(ICASSP),Toulouse,Frace,2006.NaiveOlieRegularizedRiskMiimizatioAlgorithm(NORMA),asproposediJ.Kivie,A.SmolaadC.Williamso."OlieLearigwithKerels",IEEETrasactiosoSigalProcessig,volume52,o.8,pages2165-2176,2004.Fixed-BudgetKerelRecursiveLeast-Squares(FB-KRLS),asproposediS.VaVaerebergh,I.Satamaria,W.LiuadJ.C.Pricipe,"Fixed-BudgetKerelRecursiveLeast-Squares",2010IEEEIteratioalCofereceoAcoustics,Speech,adSigalProcessig(ICASSP2010),Dallas,Texas,U.S.A.,March2010.IcompleteCholeskyDecompositio(ICD),asproposediFracisR.BachadMichaelI.Jorda."KerelIdepedetCompoetAalysis",JouralofMachieLearigResearch,volume3,pages1-48,2002.KerelRecursiveLeast-SquaresTracker(KRLS-T),asproposediM.Lazaro-Gredilla,S.VaVaereberghadI.Satamaria,"ABayesiaApproachtoTrackigwithKerelRecursiveLeast-Squares",2011IEEEIteratioalWorkshopoMachieLearigforSigalProcessig(MLSP2011),Beijig,Chia,September,2011.KerelCaoicalCorrelatioAalysis(KCCA),asproposediD.R.Hardoo,S.SzedmakadJ.Shawe-Taylor,"CaoicalCorrelatioAalysis:AOverviewwithApplicatiotoLearigMethods",NeuralComputatio,Volume16(12),Pages2639--2664,2004.QuatizedKerelLeastMeaSquares(QKLMS),asproposediCheB.,ZhaoS.,ZhuP.,PricipeJ.C."QuatizedKerelLeastMeaSquareAlgorithm,"IEEETrasactiosoNeuralNetworksadLearigSystems,vol.23,o.1,Ja.2012,pages22-32.AlteratigKerelCaoicalCorrelatioAalysisforblidequalizatioofsigle-iputmultiple-outputWieersystems,asproposediS.VaVaerebergh,J.ViaadI.Satamaria,"BlidIdetificatioofSIMOWieerSystemsbasedoKerelCaoicalCorrelatioAalysis",acceptedforpublicatioiIEEETrasactiosoSigalProcessig,2013.Kereldesityestimatio(KDE).Kerel-basedIdetificatioofHammersteisystems(KIHAM),asproposediS.VaVaereberghadL.A.Azpicueta-Ruiz,"Kerel-BasedIdetificatioofHammersteiSystemsforNoliearAcousticEcho-Cacellatio",2014IEEEIteratioalCofereceoAcoustics,Speech,adSigalProcessig(ICASSP),Florece,Italy,May2014.Nystrommethodbasedkerelmatrixdecompositioadkerelridgeregressio,asproposediC.K.I.WilliamsadM.Seeger,"UsigtheNyströmmethodtospeedupkerelmachies."Proceedigsofthe14thAualCofereceoNeuralIformatioProcessigSystems.No.EPFL-CONF-161322.2001.HowtocotributecodetothetoolboxOptio1:emailittome(steve@gtas.dicom.uica.es)
Optio2:forkthetoolboxoGitHub,pushyourchagetoaamedbrach,thesedmeapullrequest.
Icludeatleastoe"demo"fileforeachalgorithm.
ChagelogHistoryofchages:
Chagesstartigv0.10aredocumeteditheGitrepository.
v0.9(2013-05-21)
iclusioofKDEcodeaddemomiorchagesv0.8(2013-02-11)
iclusioofAKCCAcodeaddemomiorchagesv0.7(2012-09-01):
iclusioofQKLMScodemiorchagesv0.6(2012-03-26):
iclusioofademoforkerelcaoicalcorrelatioaalysis(KCCA)v0.5(2012-02-14):
iclusioofKRLS-Tadditioofafileidetifiertoeachfilev0.4(2011-05-04):
iclusioofNORMA,fixed-budgetKRLS,kerelPCA,icompleteCholeskydecompositioiclusiooficompletecholeskydecompositioalgorithm(km_kerel_icd).icludedalistigofdepedeciesifuctioheaders.formatchage:dafaultformatfordatamatricesisowoedatapoitperrow(isteadofoepercolum).formatchage:oeiputargumetlessforoliealgorithmsv0.3(2010-12-03):
modificatiostoALD-KRLSimplemetatio.v0.2(2010-11-08):
iclusioofkerelrecursiveleast-squaresalgorithms(km_krls):ALD-KRLS(ApproximateLiearDepedecyKRLS),SW-KRLS(Slidig-WidowKRLS).correctioofmiordetailsv0.1(2010-09-08):
origialpackage,icludesliearPCAadkerelridgeregressioalgorithms.
评论