LocallyAdaptiveLabelSmoothingforPredictiveChurnDaraBahri1HeinrichJiang1Abstractaswewillshowlaterintheexperiments.Thus,evenanoptimizedtrainingprocedurecanleadtohighpredictionTrainingmodernneuralnetw...
UnlabelledDataImprovesBayesianUncertaintyCalibrationunderCovariateShiftAlexJ.Chan1AhmedM.Alaa2ZhaozhiQian1MihaelavanderSchaar123Abstracttityoflabelledi.i.ddataareavailable(Krizhevskyetal.,2012).How...
MomentumImprovesNormalizedSGDAshokCutkosky12HarshMehta1AbstractthatSGD’sconvergencerateisindependentofthedimen-sionoftheproblem,allowingittoscalemoreeasilytoWeprovideanimprovedanalysisofnormalized...
FeatureQuantizationImprovesGANTrainingYangZhao1ChunyuanLi2PingYu1JianfengGao2ChangyouChen1AbstractTrainingGANsisanotoriouslychallengingtask,asitin-volvesoptimizinganon-convexproblemforitsNashequi-T...