论文阅读报告

论文阅读报告撰写人：张胜时间：10月9号一、标题ParallelSpectralClusteringinDistributedSystems二、出处IEEETRANSACTIONSONPATTERNANALYSISANDMACHINEINTELLIGENCE三、摘要Spectralclusteringalgorithmshavebeenshowntobemoreeffectiveinfindingclustersthansometraditionalalgorithms,suchask-means.However,spectralclusteringsuffersfromascalabilityprobleminbothmemoryuseandcomputationaltimewhenthesizeofadatasetislarge.Toperformclusteringonlargedatasets,weinvestigatetworepresentativewaysofapproximatingthedensesimilaritymatrix.WecompareoneapproachbysparsifyingthematrixwithanotherbytheNystrommethod.Wethenpickthestrategyofsparsifyingthematrixviaretainingnearestneighborsandinvestigateitsparallelization.Weparallelizebothmemoryuseandcomputationondistributedcomputers.Throughanempiricalstudyonadocumentdatasetof193,844instancesandaphotodataset。四、研究的问题五、研究的目标六、研究的方法七、文章结论八、备注注：这部分主要是你对这篇文章的看法，哪些方面是值得借鉴的？哪些地方可能值得继续深入下去（改进）？