基于稀疏保持的典型相关分析及其应用.pdf

《基于稀疏保持的典型相关分析及其应用.pdf》由会员上传分享，免费在线阅读，更多相关内容在工程资料-天天文库。

1、¥ã©aÒTP391Ø©?Ò102871613-S067Æ©aÒ081200a¬ÆØ©ÄuDÕ±;.'©Û9ÙA^ïÄ)6¶yÆ!;OÅÆEâïĪ£OA^ÜrÇH®ÊÊUÆïÄ)OÅÆEâÆocNanjingUniversityofAeronauticsandAstronauticsTheGraduateSchoolCollegeofComputerScienceandTechnologyCanonicalCorrelationAnalysisBasedonSparsityPreservingwithApplicationsATh

2、esisinComputerScienceandTechnologybyZuChenAdvisedbyProf.ZhangDaoqiangSubmittedinPartialFulfillmentoftheRequirementsfortheDegreeofMasterofEngineeringDecember,2012«ìÖ<(²¤¥a¬ÆØ©´<3e?1ïÄó9ïĤJ"Ø©¥AO±I5Ú/§Ø©¥Ø¹Ù¦<®²uL½>LïĤJ§Ø¹¼H®ÊÊUƽ٦ÅƽyÖ¦^Lá"<ÇH®ÊÊUƱòÆØ©ܽÜ

3、©SN?k'êâ¥?1u¢§±æ^K

4、^ØÓÀãm'&EÚpÖ&E§<Ï~©O3ùüÀãþ?1Õá!OÔö©aì§,23Õáêâ8þy©a.´Äk"ù{==¦^ØÓÀãpÖ&E§¿vkòØÓÀãm'&E

5、^å5";.'©ÛõÚO{±ÏLïÄü

6、õCþm5'X§éõÀãêâ?1AJ"d§DÕL«®¤cÅìÆSÚª£OïÄ9:"¦^

7、DÕL«±ü$$E,ݧ!;m"©òDÕL«gÚ;.'©Û§JÑ«kiÒõÀãÆS{"ÌïÄóXeµÄk§JÑÄuDÕ±;.'©Û{"ÏL¦^DÕ±g·AÀJÛÜ&E§ÓÏL3ÓamOݧòaO&Ei{¥§¦¤kiÒÆS{",§Ú'g§OrØÓÀãؤéméX"ÏL¦^ØE

8、ò{í2Am§¦{U?nE,5¯K"duÚaO&EÚ'§ÄuDÕ±;.'©ÛØ=±J'A§±3ü3¦UõO&E"3<óêâ8!õAÃNêâ8ÚORL§PIE§YALEn<òêâ8þ?1é'¢§y{

9、k5"Ùg§/ÏaS'Úam'g§òÄuDÕ±;.'©Û?1í2§¦ÙU?n"¹"3<óêâ8!õAÃNêâ8ÚYALE<òêâ8þ?1X¢§¿3Æãþ?1õ©a¢"l¢¥·uy§·{Ø=UkJA?1©aé"êدa"'cµõÀãÆS§;.'©Û§AJ§DÕ±§"iÄuDÕ±;.'©Û9ÙA^ABSTRACTWiththedevelopmentofscienceandtechnology,itisgettingeasiertogetsamedatafromanumberofdifferentviews

10、,whilethetraditionalmachinelearningmethodsmostlyworkonaviewonly.Inordertotakeadvantageofthecorrelationandcomplementaryinformationofdifferentviews,peopleusuallytrainclassifiersindependentlyanditeratively,andthenverifywhethertheclassificationmodelsonindependentdatasetsareeffective.Howe

11、ver,thesemethodsusethecomplementaryinformationonlyinsteadofboth.Asoneofmultivariatestatisticalmethods,Canonicalcorrelationanalysis(CCA)canrevealthelinearrelationshipbetweentwogroupsofmultivariatedataformulti-viewfeatureextraction.What’smore,assparserepresentationhasbeenahotspotofma

12、chinelearningandpatternrec