[NIPS 2012] Projection Retrieval for Classification

[NIPS 2012] Projection Retrieval for Classification

ID:40254155

大小:416.22 KB

页数:9页

时间:2019-07-29

[NIPS 2012] Projection Retrieval for Classification_第1页
[NIPS 2012] Projection Retrieval for Classification_第2页
[NIPS 2012] Projection Retrieval for Classification_第3页
[NIPS 2012] Projection Retrieval for Classification_第4页
[NIPS 2012] Projection Retrieval for Classification_第5页
资源描述:

《[NIPS 2012] Projection Retrieval for Classification》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库

1、ProjectionRetrievalforClassificationMadalinaFiterauArturDubrawskiMachineLearningDepartmentSchoolofComputerScienceCarnegieMellonUniversityCarnegieMellonUniversityPittsburgh,PA15213Pittsburgh,PA15213mfiterau@cs.cmu.eduawd@cs.cmu.eduAbstractInmanyapplications,classificationsystemsoft

2、enrequirehumaninterventionintheloop.Insuchcasesthedecisionprocessmustbetransparentandcomprehen-sible,simultaneouslyrequiringminimalassumptionsontheunderlyingdatadis-tributions.Totacklethisproblem,weformulateanaxis-alignedsubspace-findingtaskundertheassumptionthatqueryspecificinfor

3、mationdictatesthecomple-mentaryuseofthesubspaces.Wedeveloparegression-basedapproachcalledRECIPthatefficientlysolvesthisproblembyfindingprojectionsthatminimizeanonparametricconditionalentropyestimator.Experimentsshowthatthemethodisaccurateinidentifyingtheinformativeprojectionsofthe

4、dataset,pickingthecorrectviewstoclassifyquerypoints,andfacilitatesvisualevaluationbyusers.1IntroductionandproblemstatementInthedomainofpredictiveanalytics,manyapplicationswhichkeephumanusersinthelooprequiretheuseofsimpleclassificationmodels.Often,itisrequiredthatatest-pointbe‘exp

5、lained’(classi-fied)usingasimplelow-dimensionalprojectionoftheoriginalfeaturespace.ThisisaProjectionRetrievalforClassificationproblem(PRC).Theinteractionwiththeuserproceedsasfollows:theuserprovidesthesystemaquerypoint;thesystemsearchesforaprojectioninwhichthepointcanbeaccuratelycl

6、assified;thesystemdisplaystheclassificationresultaswellasanillustrationofhowtheclassificationdecisionwasreachedintheselectedprojection.SolvingthePRCproblemisrelevantinmanypracticalapplications.Forinstance,consideranuclearthreatdetectionsysteminstalledatabordercheckpoint.Vehiclescro

7、ssingtheborderarescannedwithsensorssothatalargearrayofmeasurementsofradioactivityandsecondarycontextualinfor-mationisbeingcollected.Theseobservationsarefedintoaclassificationsystemthatdetermineswhetherthescannedvehiclemaycarryathreat.Giventhepotentiallydevastatingconsequencesofaf

8、alsenegative,abordercontrolagentisrequestedtova

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。