资源描述：
《Visual Clustering in Parallel Coordinates》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、Eurographics/IEEE-VGTCSymposiumonVisualization2008Volume27(2008),Number3A.Vilanova,A.Telea,G.Scheuermann,andT.Möller(GuestEditors)VisualClusteringinParallelCoordinatesHongZhou1,XiaoruYuan2,HuaminQu1,WeiweiCui1,BaoquanChen31ComputerScience&EngineeringDepartment,TheHongKo
2、ngUniversityofScienceandTechnology,HongKong2SchoolofEECS&KeyLaboratoryofMachinePerception(MinistryofEducation),PekingUniversity,China3ShenzhenInstituteofAdvancedTechnology,ChineseAcademyofSciences,ChinaAbstractParallelcoordinateshavebeenwidelyappliedtovisualizehigh-dime
3、nsionalandmultivariatedata,discerningpatternswithinthedatathroughvisualclustering.However,theeffectivenessofthistechniqueonlargedataisreducedbyedgeclutter.Inthispaper,wepresentanovelframeworktoreduceedgeclutter,consequentlyimprovingtheeffectivenessofvisualclustering.Wee
4、xploitcurvededgesandoptimizethearrangementofthesecurvededgesbyminimizingtheircurvatureandmaximizingtheparallelismofadjacentedges.Theoverallvisualclusteringisimprovedbyadjustingtheshapeoftheedgeswhilekeepingtheirrelativeorder.Theexperimentsonseveralrepresentativedatasets
5、demonstratetheeffectivenessofourapproach.Keywords:Parallelcoordinates,visualclustering,multi-methodshavebeenproposed,whilemostofthemareonlyvariatedatavisualization,clutterreduction.goodforcertainkindsofdata.Inaddition,itisdifficultforuserstocontroltheclusteringlevelthrou
6、ghsettingparame-terswithoutexpertknowledge.1.IntroductionUnderstandingcomplexhigh-dimensionaldatasetsisanim-Insteadofusingdata-centricclustering,inthispaperweportantyetchallengingproblem.Amongvarioustechniquesproposetoperformvisualclusteringbygeometricallyde-developed,p
7、arallelcoordinates[ID90]havebeenwidelyformingandgroupingpolylineswhiletheyarebeingplot-adoptedforthevisualizationofhigh-dimensionalandmul-ted.Ourclusteringisachievedbyanalyzingthegeometrictivariatedatasets.Byusingparallelaxesfordimensions,therelationshipbetweenpolylines
8、ratherthanthedataitself.AnparallelcoordinatestechniquecanrepresentN-dimensionaloptimizationschemeisdesignedtom
