WhatDoesRotationPredictionTellUsaboutClassifierAccuracyunderVaryingTestingEnvironments?WeijianDeng1StephenGould1LiangZheng1AbstractThisworkinvestigatesaninterestingproblem:howtoes-timateclassifiera...
WhenDoesDataAugmentationHelpWithMembershipInferenceAttacks?YigitcanKaya1TudorDumitras,1Abstractetal.,2018;Shokrietal.,2017).Inparticular,thisenablesmembershipinferenceattacks(MIAs)tofindoutwhetherD...
HowDoesLossFunctionAffectGeneralizationPerformanceofDeepLearning?ApplicationtoHumanAgeEstimationAliAkbari1MuhammadAwais1ManijehBashar2JosefKittler1Abstractconditions,cameraquality,headpose,makeupap...
DoestheMarkovDecisionProcessFittheData:TestingfortheMarkovPropertyinSequentialDecisionMakingChengchunShi1RunzheWan2RuiSong2WenbinLu2LingLeng3Abstract1.1.ContributionsandadvancesofourtestTheMarkovas...
DoesLabelSmoothingMitigateLabelNoise?MichalLukasik1SrinadhBhojanapalli1AdityaKrishnaMenon1SanjivKumar1AbstractInterestingly,therearetwocompetingintuitions.Ontheonehand,smoothingmightmitigatetheprob...
HowDoesDisagreementHelpGeneralizationagainstLabelCorruption?XingruiYu1BoHan2JiangchaoYao3GangNiu2IvorW.Tsang1MasashiSugiyama24Abstractthesecomplexmodelscanfullymemorizenoisylabels(Zhangetal.,2017;A...
DoesDataAugmentationLeadtoPositiveMargin?ShashankRajput1ZhiliFeng1ZacharyCharles2Po-LingLoh3DimitrisPapailiopoulos2AbstractaugmentedImageNetusingtranslations,horizontalreflec-tions,andalteredintens...
DoesDistributionallyRobustSupervisedLearningGiveRobustClassifiers?WeihuaHu12GangNiu2IsseiSato12MasashiSugiyama21AbstractDistributionallyRobustSupervisedLearning(DRSL)isapromisingparadigmtotacklethi...
AnAlternativeView:WhenDoesSGDEscapeLocalMinima?RobertKleinberg1YuanzhiLi2YangYuan1Abstractmoreiterationstoconverge,butfewergradientevaluationsperiteration.Therefore,forthestandardempiricalriskmin-S...