健壮的流形学习算法及其应用研究.pdf

q ● ，11．一I ■ ≮ 摘要 …7 洲Y…ii■1 ㈣8㈣0 洲0洲7Ⅲ7 Abstract dataintherealworldale oftenof direct of hi吐dimension，the Many application traditionaldata willfacethe of”curseof miningtechnology problem dimensionality”． researchoneffectivedimensionreduction fordata ona Therefore，the algorithmanalysis low-dimensionalbecomesa traditional space hotspot．However,manydimensionality reduction focus onthelineardistributed cannotmeet algorithmsonly data,which the needsofmassivenonlineardistributeddatainrealworld．Asanonlinear analysis reduction wellinthedomain tool，manifold dimensionality leamingalgorithmsperform ofvisualizationandclassificationfornonlinear have usedin data,andbeen widely many areasof manifold still havesome society．However,theexisting learningalgorithms the to becomesaboRleneckof noise，which shortcomings，particularlysensitivity manifold that blockitfrombeen learningalgorithmsmay widely researchonrobustmanifold becomesanimmediateneed．’ learnin