一种新的图像置乱算法

《一种新的图像置乱算法》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、计　算　机　研　究　与　发　展　　　　　　　　　　　　　　　　2016年（注：此模板非完整论文，已做删减。只看格式，勿参考内容）异质网中基于张量表示的动态离群点检测方法题目三号刘露1左万利1,2彭涛1,2作者四号宋体，作者和单位的对应关系标注在作者姓名的右上角1（吉林大学计算机科学与技术学院长春130012）单位小五号2（符号计算与知识工程教育部重点实验室（吉林大学）长春130012）1（liulu12@mails.jlu.edu.cn）小五号TensorRepresentationBasedDynamicOutlie

2、rDetectionMethodinHeterogeneousNetworkTitle四号LiuLu1,ZuoWanli1,2,andPengTao1,2Name五号1（CollegeofComputerScienceandTechnology,JilinUniversity,Changchun130012）Depart.Correspond小五号2（KeyLaboratoryofSymbolComputationandKnowledgeEngineering（JilinUniversity）,MinistryofEd

3、ucation,Changchun130012）AbstractMiningrichsemanticinformationhiddeninheterogeneousinformationnetworkisanimportanttaskindatamining.Thevalue,datadistributionandgenerationmechanismofoutliersarealldifferentfromthatofnormaldata.Itisofgreatsignificanceofanalyzingitsge

4、nerationmechanismoreveneliminatingoutliers.Outlierdetectioninhomogeneousinformationnetworkhasbeenstudiedandexploredforalongtime.However,fewofthemareaimingatdynamicoutlierdetectioninheterogeneousnetworks.Manyissuesneedtobesettled.Duetothedynamicsoftheheterogeneou

5、sinformationnetwork,normaldatamaybecomeoutliersovertime.ThispaperproposesadynamicTensorRepresentationBasedOutlierdetectionmethod,calledTRBOutlier.Itconstructstensorindextreeaccordingtothehighorderdatarepresentedbytensor.Thefeaturesareaddedtodirectitemsetandindir

6、ectitemsetrespectivelywhensearchingthetensorindextree.Meanwhile,wedescribeaclusteringmethodbasedonthecorrelationofshorttextstojudgewhethertheobjectsindatasetschangetheiroriginalclustersandthendetectoutliersdynamically.Thismodelcankeepthesemanticrelationshipinhet

7、erogeneousnetworksasmuchaspossibleinthecaseoffullyreducingthetimeandspacecomplexity.Theexperimentalresultsshowthatourproposedmethodcandetectoutliersdynamicallyinheterogeneousinformationnetworkeffectivelyandefficiently.Abstract五号，至少200字，影响EI索引Keywordsdynamicoutli

8、erdetection;heterogeneousinformationnetwork;tensorrepresentation;tensorindextree;clusteringKeywords五号摘要挖掘隐藏在异质信息网络中丰富的语义信息是数据挖掘的重要任务之一．离群点在值、数据分布、和产生机制上都明显不同于正常数据对象．检