WhatMakesforEnd-to-EndObjectDetection?PeizeSun1YiJiang2EnzeXie1WenqiShao3ZehuanYuan2ChanghuWang2PingLuo1AbstractFigure1.End-to-endobjectdetection.Non-end-to-endobjectdetectorsrequireNMStoremoveredu...
What’sintheBox?ExploringtheInnerLifeofNeuralNetworkswithRobustRulesJonasFischer1AnnaOla´h1JillesVreeken2AbstractYaretypicallyactivewhenneuronsXare.Forrobustnessweexplicitlyallowfornoise,andtoensu...
WhatAreBayesianNeuralNetworkPosteriorsReallyLike?PavelIzmailov1SharadVikram2MatthewD.Hoffman2AndrewGordonWilson1Abstractforneuralnetworkspromisesimprovedpredictions,reli-ableuncertaintyestimates,an...
WhatDoesRotationPredictionTellUsaboutClassifierAccuracyunderVaryingTestingEnvironments?WeijianDeng1StephenGould1LiangZheng1AbstractThisworkinvestigatesaninterestingproblem:howtoes-timateclassifiera...
HierarchicalVAEsKnowWhatTheyDon’tKnowJakobD.Havtorn12JesFrellsen1SørenHauberg1LarsMaaløe12AbstractFigure1.ReconstructionsusingahierarchicalVAEtrainedonFashionMNIST.ReconstructionqualityofOODdata...
DecidingWhattoLearn:ARate-DistortionApproachDilipArumugam1BenjaminVanRoy1Abstractofinformationthat,whileinsufficienttofullyidentifytheenvironment,sufficestoguideeffectivedecisions.Then,Agentsthatle...
LearningWhattoDeferforMaximumIndependentSetsSungsooAhn1YounggyoSeo1JinwooShin1Abstractpracticalsituations.Overthepastdecades,researchershavemadesignificanteffortsforresolvingthisissuebydesign-Desig...
KnowingTheWhatButNotTheWhereinBayesianOptimizationVuNguyen1MichaelAOsborne1AbstractBayesianoptimizationfindstheglobalmaximizerx∗=argmaxx∈Xf(x)oftheblack-boxfunctionfbyincorporat-Bayesianoptimizat...
It’sNotWhatMachinesCanLearn,It’sWhatWeCannotTeachGalYehuda1MosheGabel2AssafSchuster1AbstractdatasetgenerationprocesscomplexityofresultingtaskCandeepneuralnetworkslearntosolveanytask,harderandinpa...
Goal-AwarePrediction:LearningtoModelWhatMattersSurajNair1SilvioSavarese1ChelseaFinn1Abstractheavysupervisionistoleveragedatathattheagentcancollectautonomouslywithoutrewardsorlabels,termedLearneddyn...
EnhancingSimpleModelsbyExploitingWhatTheyAlreadyKnowAmitDhurandhar1KarthikeyanShanmugam1RonnyLuss1Abstract1.IntroductionTherehasbeenrecentinterestinimprovingper-Simplemodelssuchasdecisiontreesorrul...
WhatistheEffectofImportanceWeightinginDeepLearning?JonathonByrd1ZacharyC.Lipton1AbstractEq[f(x)],importancesamplingproducesanunbiasedesti-matebyweightingeachsamplexaccordingtothelikelihoodImportanc...