DecompositionofUncertaintyinBayesianDeepLearningforEfficientandRisk-sensitiveLearningStefanDepeweg12Jose´MiguelHerna´ndez-Lobato3FinaleDoshi-Velez4SteffenUdluft1AbstractInthisworkweshowhowtoperfo...
DCFNet:DeepNeuralNetworkwithDecomposedConvolutionalFiltersQiangQiu1XiuyuanCheng1RobertCalderbank1GuillermoSapiro1AbstractFigure1.InaDCFNet,anL×L×M×Mconvolutionallayerisdecomposedintotheproductof...
ContextNet:DeeplearningforStarGalaxyClassificationNobleKennamer1DavidKirkby2AlexIhler1JavierSa´nchez2AbstractturestomappingsmallobjectsintheSolarSystemsuchasnear-Earthasteroids(LSSTScienceCollabor...
ContextualGraphMarkovModel:ADeepandGenerativeApproachtoGraphProcessingDavideBacciu1FedericoErrica1AlessioMicheli1AbstractMarkovModel(Bacciuetal.,2012a)andtheconstructiveapproachofNeuralNetworkforGr...
ConstrainingtheDynamicsofDeepProbabilisticModelsMarcoLorenzi1MaurizioFilippone2Abstractmenttoprovideamorepreciseandrealisticdescriptionofnaturalphenomena.Forexample,monotonicityoftheinter-Weintrodu...
ComparingDynamics:DeepNeuralNetworksversusGlassySystemsMarcoBaity-Jesi1LeventSagun23MarioGeiger3StefanoSpigler32Ge´rardBenArous4ChiaraCammarota5YannLeCun467MatthieuWyart3GiulioBiroli28Abstractdere...
CanDeepReinforcementLearningSolveErdos-Selfridge-SpencerGames?MaithraRaghu12AlexIrpan1JacobAndreas3RobertKleinberg2QuocLe1JonKleinberg2Abstractbehaviorisdifficult.Optimalbehaviorintheseenviron-ment...
BoundingandCountingLinearRegionsofDeepNeuralNetworksThiagoSerra1ChristianTjandraatmadja1SrikumarRamalingam2Abstract1010WeinvestigatethecomplexityofDeepneuralNumberoflinearregions108networks(DNN)tha...
BeyondFiniteLayerNeuralNetworks:BridgingDeepArchitecturesandNumericalDifferentialEquationsYipingLu1AoxiaoZhong2QuanzhengLi234BinDong564Abstractswhilemaintainingasimilarperformance.Thiscanbeexplaine...
BayesianUncertaintyEstimationforBatchNormalizedDeepNetworksMattiasTeye12HosseinAzizpour1KevinSmith13Abstractorifthenetworkisconfrontedwithadversarialexamples(Goodfellowetal.,2014).Whenexposedtodata...
Attention-basedDeepMultipleInstanceLearningMaximilianIlse1JakubM.Tomczak1MaxWelling1Abstractmodelthatpredictsabaglabel,e.g.,amedicaldiagnosis.Anadditionalchallengeistodiscoverkeyinstances(Liuetal.,...
AnOptimalControlApproachtoDeepLearningandApplicationstoDiscrete-WeightNeuralNetworksQianxiaoLi1ShujiHao1Abstractversionsoftheabovealgorithms.Morebroadly,necessaryconditionsforoptimalitycanbederived...
AccurateUncertaintiesforDeepLearningUsingCalibratedRegressionVolodymyrKuleshov12NathanFenner2StefanoErmon1AbstractSales(units)Raw90%ConfidenceIntervalfromaBayesianNeuralNetworkMethodsforreasoningun...
ACloserLookatMemorizationinDeepNetworksDevanshArpit12StanisławJastrze˛bski3NicolasBallas12DavidKrueger12EmmanuelBengio4MaxinderS.Kanwal5TeganMaharaj16AsjaFischer7AaronCourville128YoshuaBengio129S...
Zero-ShotTaskGeneralizationwithMulti-TaskDeepReinforcementLearningJunhyukOh1SatinderSingh1HonglakLee12PushmeetKohli3AbstractFigure1:Exampleof3Dworldandinstructions.Theagentistaskedtoexecutelongerse...
VariationalDropoutSparsifiesDeepNeuralNetworksDmitryMolchanov12ArseniiAshukha34DmitryVetrov31Abstractprobabilitiesofdroppinganeuron.Thedropoutratesaretypicallyoptimizedusinggridsearch.Toavoidtheex-...
UnimodalProbabilityDistributionsforDeepOrdinalClassificationChristopherBeckham1ChristopherPal1Abstractotherclassesmaynotnecessarilymakesenseincontext.WepresentanexampleofthisinFigure1,showingthreeP...
TowardsK-means-friendlySpaces:SimultaneousDeepLearningandClusteringBoYang1XiaoFu1NicholasD.Sidiropoulos1MingyiHong2Abstract1.IntroductionMostlearningapproachestreatdimensionalityClusteringisoneofth...
TheLossSurfaceofDeepandWideNeuralNetworksQuynhNguyen1MatthiasHein1Abstractdoesnotencounterproblemswithsuboptimallocalmin-ima.However,astheauthorsadmitthemselvesin(Good-Whiletheoptimizationproblembe...
StabilisingExperienceReplayforDeepMulti-AgentReinforcementLearningJakobFoerster1NantasNardelli1GregoryFarquhar1TriantafyllosAfouras1Philip.H.S.Torr1PushmeetKohli2ShimonWhiteson1Abstractmulti-agents...