UnitaryBranchingPrograms:LearnabilityandLowerBoundsFidelErnestoD´ıazAndino1MariaKokkou2MateusdeOliveiraOliveira3FarhadVadiee3AbstractdepthBooleancircuitscanbecomputedbywidth-5branch-ingProgramsof...
TensorProgramsIV:FeatureLearninginInfinite-WidthNeuralNetworksGregYang1EdwardJ.Hu23AbstractFigure1.PCAofWord2VecembeddingsoftopUScitiesandstates,forNTK,width-64,andwidth-∞featurelearningnetworksAs...
TensorProgramsIIb:ArchitecturalUniversalityofNeuralTangentKernelTrainingDynamicsGregYang1EtaiLittwin2AbstractIn(Yang,2020a),theNTKINITpropertywasproventoholdforstandardarchitectures,meaninganycompo...
ProbabilisticProgramswithStochasticConditioningDavidTolpin1YuanZhou2TomRainforth3HongseokYang4Abstracttheprobability(massordensity)oftheobservationgiventhedistributionisusedininference(Carpentereta...
PrincipalComponentHierarchyforSparseQuadraticProgramsRobbieVreugdenhil1VietAnhNguyen23ArminEftekhari4PeymanMohajerinEsfahani1Abstractspecifically,weconsidertheproblemWeproposeanovelapproximationhie...
HowcouldNeuralNetworksunderstandPrograms?DinglanPeng1ShuxinZheng2YataoLi2GuolinKe2DiHe2Tie-YanLiu2Abstract1.IntroductionSemanticunderstandingofProgramsisafunda-Modernsoftwaretypicallycontainstonsof...
AutomaticReparameterisationofProbabilisticProgramsMariaI.Gorinova1DaveMoore2MatthewD.Hoffman2Abstractparameterisation(CP).Ifweinsteadworkwithanauxil-iary,standardnormalvariablez˜∼N(0,1),andobtain...
SynthesizingProgramsforImagesusingReinforcedAdversarialLearningYaroslavGanin1TejasKulkarni2IgorBabuschkin2S.M.AliEslami2OriolVinyals2AbstractFigure1.SPIRALtakesasinputeitherrandomnoiseorimagesandit...
SoundAbstractionandDecompositionofProbabilisticProgramsStevenHoltzen1GuyVandenBroeck1ToddMillstein1Abstractindependencesthatcanspeedupinferencebyfactorization;forexample,low-widthtreedecompositions...
DifferentiableProgramswithNeuralLibrariesAlexanderL.Gaunt1MarcBrockschmidt1NateKushman1DanielTarlow2Abstractexplorepotentialadvantagesofthisclassofhybridmodeloverpurelyneuralsystems,withaparticular...