简介：在这份报纸，调用模糊散开地图(频分多路复用)的一个新奇方法被建议评估动画片类似，它对动画片识别，动画片聚类并且动画片再使用的应用批评。我们发现从异构的来源的特征在动画片类似评价上有不同影响。以便考虑所有特征，一种模糊一致关系被介绍把特征的偏爱顺序变换成偏爱度，重量从是计算的。基于特征和重量，摆平的差别(L2)的和能在任何动画片数据之间被计算。然而，它在一些研究工作被表明了动画片数据集躺在一低维歧管，在哪个L2距离不能直接评估类似。不同于在Isomap，在局部地线性的嵌入保存的本地附近的关系，和附近的点的类似在拉普拉斯算符Eigenmaps保存了的本地人保存的全球测地学的距离，我们采用的散开地图保存散开距离在连接二个数据的长度的所有路径上加。作为后果，这散开距离对噪音不安很柔韧。我们在使用接收装置操作弄弯的动画片分类的实验在重量赋值上显示出模糊一致关系的优秀表演。动画片类似评估上的频分多路复用表演在动画片识别的实验上被测试并且聚类。结果证明频分多路复用能与另外的方法相比更精确并且稳定地评估动画片类似。

简介：

简介：

简介：

简介：Algorithmsusedindataminingandbioinformaticshavetodealwithhugeamountofdataefficiently.Inmanyapplications,thedataaresupposedtohaveexplicitorimplicitstructures.Todevelopefficientalgorithmsforsuchdata,wehavetoproposepossiblestructuremodelsandtestifthemodelsarefeasible.Hence,itisimportanttomakeacompactmodelforstructureddata,andenumerateallinstancesefficiently.Therearefewgraphclassesbesidestreesthatcanbeusedforamodel.Inthispaper,weinvestigatedistance-hereditarygraphs.Thisclassofgraphsconsistsofisometricgraphsandhencecontainstreesandcographs.First,acanonicalandcompacttreerepresentationoftheclassisproposed.Thetreerepresentationcanbeconstructedinlineartimebyusingprefixtrees.Usually,prefixtreesareusedtomaintainasetofstrings.Inouralgorithm,theprefixtreesareusedtomaintaintheneighborhoodofvertices,whichisanewapproachunlikethelexicographicallybreadth-firstsearchusedinotherstudies.Basedonthecanonicaltreerepresentation,efficientalgorithmsforthedistance-hereditarygraphsareproposed,includinglineartimealgorithmsforgraphrecognitionandgraphisomorphismandanefficientenumerationalgorithm.Anefficientcodingforthetreerepresentationisalsopresented;itrequires[3.59n]bitsforadistance-hereditarygraphofnverticesand3nbitsforacograph.Theresultsofcodingimprovepreviouslyknownupperbounds(bothare2~(O(nlogn)))ofthenumberofdistance-hereditarygraphsandcographsto2~([3.59n])and2~(3n),respectively.

简介：BymeansofdirectanalysisoftheconnectionbetweenLoopsubdivisionsurfaceanditscontrolmeshandthecomputationofthebasisfunctions,weobtainaboundonthedistancebetweenLoopsubdivisionsurfacepatchanditscontrolmesh.Theboundcanbeusedtocomputethenumbersofsubdivisionforagiventolerance.Finally,twoexamplesarelistedinthispapertodemonstratetheapplicationsofthebound.

简介：

简介：

简介：

简介：

简介：Wepresentthesolidmodeleditdistance(SMED),apowerfulandflexibleparadigmforexploitingshapesimilaritiesamongstCADmodels.ItisdesignedtomeasurethemagnitudeofdistortionsbetweentwoCADmodelsinboundaryrepresentation(B-rep).Wegivetheformaldefinitionbyanalogywithgrapheditdistance,oneofthemostpopulargraphmatchingmethods.Toavoidtheexpensivecomputationalcostpotentiallycausedbyexactcomputation,anapproximateprocedurebasedonthealignmentoflocalstructuresetsisprovidedinaddition.Inordertoverifytheflexibility,wemakeintensiveinvestigationsonthreetypicalapplicationsinmanufacturingindustry,anddescribehowourmethodcanbeadaptedtomeetthevariousrequirements.Furthermore,amultilevelmethodisproposedtomakefurtherimprovementsofthepresentedalgorithmonbotheffectivenessandefficiency,inwhichthemodelsarehierarchicallysegmentedintotheconfigurationsoffeatures.ExperimentresultsshowthatSMEDservesasareasonablemeasurementofshapesimilarityforCADmodels,andtheproposedapproachprovidesremarkableperformanceonareal-worldCADmodeldatabase.

简介：

简介：