SoftthenHard:RethinkingtheQuantizationinNeuralImageCompressionZongyuGuo1ZhizhengZhang1RunsenFeng1ZhiboChen1AbstractQuantizationisoneofthekeychallengesforneuralimagecompression.Sincethegradientofqua...
Low-PrecisionReinforcementLearning:RunningSoftActor-CriticinHalfPrecisionJohanBjorck1XiangyuChen1ChristopherDeSa1CarlaP.Gomes1KilianQ.Weinberger1Abstractlearning,anemergingtrendforacceleratingdeepl...
ConViT:ImprovingVisionTransformerswithSoftConvolutionalInductiveBiasesSte´phaned’Ascoli12HugoTouvron2MatthewL.Leavitt2AriS.Morcos2GiulioBiroli12LeventSagun2Abstract1.IntroductionConvolutionalarch...
SoftThresholdWeightReparameterizationforLearnableSparsityAdityaKusupati1VivekRamanujan2RaghavSomani1MitchellWortsman1PrateekJain3ShamKakade1AliFarhadi1Abstractbottleneckinthereal-worlddeploymentoft...
Operation-AwareSoftChannelPruningusingDifferentiableMasksMinsooKang1BohyungHan1AbstractNotUpdated1.00.0UpdatedstrongweakBinaryMasksFeatureMapWeproposeasimplebuteffectivedata-drivenDifferentiableMas...
AnalyzingandImprovingRepresentationswiththeSoftNearestNeighborLossNicholasFrosst1NicolasPapernot1GeoffreyHinton1AbstracttheSoftnearestneighborloss(Salakhutdinov&Hinton,2007),whichweexploretomeasure...
SoftActor-Critic:Off-PolicyMaximumEntropyDeepReinforcementLearningwithaStochasticActorTuomasHaarnoja1AurickZhou1PieterAbbeel1SergeyLevine1Abstractnetworksholdsthepromiseofautomatingawiderangeofdeci...