成果matlab六課教程_第1頁
成果matlab六課教程_第2頁
成果matlab六課教程_第3頁
成果matlab六課教程_第4頁
成果matlab六課教程_第5頁
已閱讀5頁,還剩67頁未讀 繼續(xù)免費閱讀

下載本文檔

版權說明:本文檔由用戶提供并上傳,收益歸屬內容提供方,若內容存在侵權,請進行舉報或認領

文檔簡介

1 1 3IntroductiontoMatlab3IntroductiontoMatlabWhatisaMatlabToolboxesarespecializedcollectionsofMatlabWhydoweneedthe4WhattypeoftoolboxesdoesMatlab4WhattypeoftoolboxesdoesMatlab??ParallelComputing數(shù)學、統(tǒng)計與優(yōu)SymbolicMathPartialDifferentialEquationStatisticsandMachineLearningCurveFittingOptimizationNeuralNetworkModel-BasedCalibrationControlSystemToolboxFuzzyLogicToolboxRobustControlModelPredictiveControlToolboxAerospaceToolboxRoboticsSystem???????5WhattypeoftoolboxesdoesMatlab5WhattypeoftoolboxesdoesMatlab信號處無線DSPSystemToolboxAudioSystemToolboxWaveletToolboxRFToolboxLTESystemToolboxWLANSystem??????????圖像處理與計算機視ImageProcessingToolboxVisionHDLToolboxMappingToolbox?????測試&DataAcquisitionToolboxImageAcquisitionToolboxOPCToolboxVehicleNetwork?????6Whattypeoftoolboxesdoes6WhattypeoftoolboxesdoesMatlab計算FinancialDatafeedToolboxDatabaseToolboxSpreadsheetLink(forMicrosoftExcel)FinancialInstrumentsToolbox???????計算生Bioinformatics代碼MATLABHDLVisionHDLHDLFilterDesignHDLFixed-Point7Whattypeoftoolboxes7WhattypeoftoolboxesdoesMatlabMATLABMATLABCompilerSpreadsheetLink(forMicrosoftMATLABProductionDatabaseMATLABReportFromDeepLearning8WherecanI8WherecanIfindinformationregarding9ClusteringusingtheStatistics9ClusteringusingtheStatisticsandMachineLearningToolboxTheMatlabStatisticsandMachineLearningToolboxhasmanyStatisticsandDataimportandexport,descriptivestatistics,visualizationProbabilityDistributionsDatafrequencymodels,randomsamplegeneration,parameterestimationHypothesisTestst-test,F-test,chi-squaregoodness-of-fittest,andClusterUnsupervisedlearningtechniquestofindnaturalgroupingsandpatternsinAnalysisofvarianceandcovariance,multivariateANOVA,repeatedmeasuresANOVALinear,generalizedlinear,nonlinear,andnonparametrictechniquesforsupervisedSupervisedlearningalgorithmsforbinaryandmulticlassproblemsDimensionalityReductionPCA,factoranalysis,nonnegativematrixfactorization,sequentialfeatureselection,andmoreIndustrialStatisticsDesignofexperiments(DOE);survivalandreliabilityanalysis;statisticalprocesscontrolSpeedUpStatisticalComputationsParallelordistributedcomputationofstatistical????????????????????WhatisAWhatisAwaytoform'naturalgroupings'inyourItisaformofunsupervisedlearning–yougenerallydoNOThaveexamplesdemonstratinghowthedatashouldbegroupedWhydoweneedMarketsegmentation:Assistmarketerstoidentifydistinctsub-groupsofcustomersinorderWhydoweneedMarketsegmentation:Assistmarketerstoidentifydistinctsub-groupsofcustomersinordertodeveloptargetedmarketingprogramsGuiltbyassoito”:IdentifyinggroupsofgenesthatbehavesimilarlyunderasetofdifferentexperimentalClusterClusterClusterClustering Clustering K-means K-means K-meansThek-meansalgorithmK-meansThek-meansalgorithmpartitionsthedataintokexclusiveFeatureFeatureK-meansThek-meansalgorithmpartitionsK-meansThek-meansalgorithmpartitionsthedataintokexclusiveK=FeatureFeatureK-meansThek-meansalgorithmpartitionsK-meansThek-meansalgorithmpartitionsthedataintokmutuallyexclusiveK=FeatureFeatureHowdoesitK-meansFormalminimizetotalintra-clusterKd(xjK-meansFormalminimizetotalintra-clusterKd(xj,iSiistheithcluster(i=1,2,...,i1xjμiistheithcentroidofallthepointsinclusterdisadistanceOptimalSuboptimalK-meansIfweknewK-meansIfweknewtheclusterassignmentofeachwecouldeasilycomputethecentroidsIfweknewthecentroidpositionswecouldassigneachpointtoaButwedon’tknowneitherofK-meansAlgorithmChooseK-meansAlgorithmChoosethenumberofclusters,RandomlychooseinitialpositionsofKAssigneachofthepointstothe“nat(dependsondistanceK=K-meansAlgorithmK-meansAlgorithmChoosethenumberofclusters-RandomlychooseinitialpositionsofKAssigneachofthepointstothe“astn”(dependsondistanceK=K-meansAlgorithmChooseK-meansAlgorithmChoosethenumberofclusters-RandomlychooseinitialpositionsofKAssigneachofthepointstothe“artend”ondistancemeasure)CalculatetheintraclusterRe-computecentroidIfsolutionconverges→K=K-meansAlgorithmChooseK-meansAlgorithmChoosethenumberofclusters-RandomlychooseinitialpositionsofKAssigneachofthepointstothe“restetrod”ondistancemeasure)CalculatetheintraclusterRe-computecentroidIfsolutionconverges→K=K-meansAlgorithmChoosethenumberK-meansAlgorithmChoosethenumberofclusters-RandomlychooseinitialpositionsofKcentroidsAssigneachofthepointstothe“restetrod”ondistancemeasure)CalculatetheintraclusterRe-computecentroidpositionsIfsolutionconverges→Stop!K=??????K-meansAlgorithmChooseK-meansAlgorithmChoosethenumberofclusters-RandomlychooseinitialpositionsofKAssigneachofthepointstothe“restetrod”ondistancemeasure)CalculatetheintraclusterRe-computecentroidIfsolutionconverges→K=K-meansAlgorithmChooseK-meansAlgorithmChoosethenumberofclusters-RandomlychooseinitialpositionsofKAssigneachofthepointstothe“artend”ondistancemeasure)CalculatetheintraclusterRe-computecentroidIfsolution(theintraclustervarianceddnchange)→K=K-means:otherthingsweneedtoHowshouldK-means:otherthingsweneedtoHowshouldwechooseWhattypeofdistancemeasurescanweuse,andhowtochoosebetweenthem?((x2–x1)2+–y1)2)Sumofabsolute|x2–x1|+|y2–1–tAndK-means:otherthingsweneedK-means:otherthingsweneedtoDoesthealgorithmconvergencetoanoptimalCanyouthinkofstrategiesforsolvingBeforewelearnhowtoBeforewelearnhowtodoK-meansinMatlablet’slookatsomerealdata…Inthe1920's,botanistscollectedmeasurementsonsepalsepalpetalpetalof150iris,50fromeachofthreespecies(setosa,versicolor,ThemeasurementsbecameknownasFisher'sirisFisher’sIrisload4Fisher’sIrisload4SampleSampleSampleSample>>1>>'versicolor''virginica','virginica','setosa','setosa','setosa','setosa',FeatureFeatureFeatureFeatureExploringcorrelationsintheExploringcorrelationsintheFisher’sIrisparam_names={'sepallength','sepalwidth','petallength','petaltext([.05.30.55.80],[-0.1,-0.1,-0.1,-0.1],param_names,text([-0.12,-0.12,-0.12,-0.12],[0.800.550.300.05],'FontSize',12,ThepetallengthandwidtharehighlyVisualizingFisher’sIris%76543218276545SepalSepalVisualizingFisher’sIris%76543218276545SepalSepalPetal4K-meansusingDoingK-meansinMatlabis=K-meansusingDoingK-meansinMatlabis=BydefaultkmeansusessquaredEuclidiandistanceTheKTheclusterbelongstoK-meansusingDisplayingthealgorithm=K-meansusingDisplayingthealgorithm=210sumof123441112=K-meansusingClusteringptsymbK-meansusingClusteringptsymb=%Plotclusterpointsfori=1:2clust=(cidx2==i);holdNoticethatclusteringisdonebutvisualization%Plotclustercentroidholdxlabel('SepalLength');ylabel('SepalWidth');zlabel('Petalgridtitle('IrisdataclusteredwithK-meanswhereK=K-meansusingClusteringCluster765Cluster43218276544SepalSepalPetal K-meansusingClusteringCluster765Cluster43218276544SepalSepalPetal K-meansusingClustering765becausetheupperclusterisspreadout,thesethreepointsareclosertothecentroidofthelowerclusteruppercluster43218276544SepalSepalPetal K-meansusingClustering765becausetheupperclusterisspreadout,thesethreepointsareclosertothecentroidofthelowerclusteruppercluster43218276544SepalSepalPetal K-meansusingIncreasingthenumberofK-meansusingIncreasingthenumberof=11112430sumdistances=123455K-meansusingClusteringK-meansusingClusteringfori=clust=(cidx3==i);holdonholdxlabel('SepalLength');ylabel('SepalWidth');zlabel('PetalgridK-meansusingK-meansusingK-meansusingAvoidinglocalminimausingareplicates=4K-meansusingAvoidinglocalminimausingareplicates=458=====K-meansusing76543218276544SepalSepalPetal K-meansusing76543218276544SepalSepalPetal K-meansusing=7Wecanusethecosfunctionasadistancemeasurebetween6543218276544SepalSepalPetal K-meansusing=7Wecanusethecosfunctionasadistancemeasurebetween6543218276544SepalSepalPetal K-meansusingWhichdistancemeasureisK-meansusingWhichdistancemeasureismoreeWeknowthelabelofeachsample.Wecancompareclustersdiscoveredbykmeanstotheactualflowertypes.Note:usuallyinunsupervisedlearningwedoNOTknowthelabelsoftheK-meansusing%TestingtheclusteringaccuracyK-meansusing%Testingtheclusteringaccuracyfori=clust=find(cidx_cos==i);holdonxlabel('Sepalylabel('SepalWidth');gridonmiss=find(cidx_cos===holdK-meansusingCosinebaseddistance:576543218276544SepalPetal K-meansusingCosinebaseddistance:576543218276544SepalPetal K-meansusingEuclideanbaseddistance:1476543218276544SepalSepalPetal K-meansusingEuclideanbaseddistance:1476543218276544SepalSepalPetal HowtochooseWeneedaquantitativemethodtoHowtochooseWeneedaquantitativemethodtoassessthequalityofaThesilhouettevalueofapointisameasureofhowsimilarapointispointsinitsownclustercomparedtopointsinother-Formaldefinition:s(i)max(a(i),istheaveragedistanceofthepointitotheotherpointsinitsownclusterd(i,C)istheaveragedistanceofthepointitotheotherpointsintheclusteristheminimald(i,C)overallclustersotherthanHowtochooseSilhouettevaluesrangesfromHowtochooseSilhouettevaluesrangesfrom-1to→~=objectiswell→~objectisontheborderbetween2→~=-ObjectisclassifiedThesilhouettecoefficientistheaveragesilhouettevalueoverItisaquantitativemeasurethatcanassessthequalityofHowtochooseToHowtochooseTodemonstratetheutilityofthesilhouettecoefficientwecantestitonsyntheticdataforwhichweknowthenumberofx1=randn(1,100);y1=randn(1,scatter(x1,y1,25,[100],holdx2=randn(1,100)+3;y2=randn(1,scatter(x2,y2,25,[010],+x3=randn(1,100)+8;y3=randn(1,100);scatter(x3,y3,25,[001],'filled');holdHowtochooseTodemonstratetheutilityofthesilhouettecoefficientwecantestitonsyntheticdataforwhichweknowthenumberofWeknowthatKis65432100HowtochooseTodemonstratetheutilityofthesilhouettecoefficientwecantestitonsyntheticdataforwhichweknowthenumberofWeknowthatKis654321002468HowtochooseWerunthek-meansalgorithmfordifferentx=[x1,x2,y=[y1,data=[x',K====K=K=6665554443332221110000505 05HowtochooseWerunthek-meansalgorithmfordifferentx=[x1,x2,y=[y1,data=[x',K====K=K=6665554443332221110000505 05Howtochoose>>[silh2,h]=>>1arepoorly201Howtochoose>>[silh2,h]=>>1arepoorly201Howtochoose>>[silh3,h]>>12301SilhouetteHowtochoose>>[silh3,h]>>12301SilhouetteHowtochoose>>[silh4,h]>>12340SilhouetteHowtochoose>>[silh4,h]>>12340Silhouette1HowtochooseOptimalSilhouettevalueisachievedwhenK=32345K6HowtochooseOptimalSilhouettevalueisachievedwhenK=32345K678MeansilhouetteK-means investigateK-means investigategroupinginyourdata,simultaneouslyoveravarietyofscalesAlgorithm1)DeterminethedistancebetweenAlgorithm1)DeterminethedistancebetweeneachpairofdifferentTypesofdistances(Euclidean,correlation,1234512345Algorithm1)DeterminethedistancebetweeneachpairAlgorithm1)Determinethedistancebetweeneachpairof2)IterativelygrouppointsintoabinaryhierarchicaltreeConnecttheclosestpairofpointsandre-computedistance9876Thedistanceatwhichthepairofpointswere34521Algorithm1)DeterminethedistancebetweeneachpairofAlgorithm1)Determinethedistancebetweeneachpairof2)Iterativelygrouppointsintoabinaryhierarchicaltree3)Cutthehierarchicaltreeinto34521Hierarchicalclustering,otherthingsweHierarchicalclustering,otherthingswetoTypesofSinglelinkageiorDistancebetweengroupsisdefinedasthebetweentheclosestpairofpointsfromeachHierarchicalclustering,otherthingsweHierarchicalclustering,otherthingswetoTypesofCompletelinkageibDistancebetweengroupsisdefinedasthedistancebetweenthemostdistantpairofpointsfromtwoHierarchicalclustering,otherHierarchicalclustering,otherthingsweneedtoconsiderTypesofAveragelinkageclustering:Thedistancebetweentwoclustersisdefinedastheaverageofdistancesbetweenallpairsofpoints(ofoppositeHierarchicalclustering,otherthingswetoHierarchicalclustering,otherthingswetoWheretocuttheCuttingatanarbitraryHierarchicalclustering,otherthingsweneedtoconsiderWheretocuttheHierarchicalclustering,otherthingsweneedtoconsiderWheretocutthe??CuttingatanarbitraryCuttingatinconsistencyComparetheheightofeachlinkinthetreewiththeheightsoflinksbelowit:IfapproximatelyequalThislinkexhibitsahighlevelofconsistency.Therearenodistinctdivisionsbetweentheobjectsjoinedatthislevelofthehierarchy.?IfheightsdifferThislinkissaidtobeinconsistentinrespecttothelinksbelowit.Thisindicatestheborderofanaturaldivisioninadataset.Forformaldefinitionsseetoolbox??HierarchicalclusteringusingLoadtheIris>>HierarchicalclusteringusingLoadtheIris>>load1)Computethedistancesbetweeneach>>euc

溫馨提示

  • 1. 本站所有資源如無特殊說明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請下載最新的WinRAR軟件解壓。
  • 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請聯(lián)系上傳者。文件的所有權益歸上傳用戶所有。
  • 3. 本站RAR壓縮包中若帶圖紙,網頁內容里面會有圖紙預覽,若沒有圖紙預覽就沒有圖紙。
  • 4. 未經權益所有人同意不得將文件中的內容挪作商業(yè)或盈利用途。
  • 5. 人人文庫網僅提供信息存儲空間,僅對用戶上傳內容的表現(xiàn)方式做保護處理,對用戶上傳分享的文檔內容本身不做任何修改或編輯,并不能對任何下載內容負責。
  • 6. 下載文件中如有侵權或不適當內容,請與我們聯(lián)系,我們立即糾正。
  • 7. 本站不保證下載資源的準確性、安全性和完整性, 同時也不承擔用戶因使用這些下載資源對自己和他人造成任何形式的傷害或損失。

評論

0/150

提交評論