LearningtoCombineTop-DownandBottom-UpSignalsinRecurrentNeuralNetworkswithAttentionoverModulesSarthakMittal1AlexLamb2AnirudhGoyal2VikramVoleti12MurrayShanahan3GuillaumeLajoie12MichaelMozer4YoshuaBen...
Divide,Conquer,andCombine:aNewInferenceStrategyforProbabilisticProgramswithStochasticSupportYuanZhou1HongseokYang2YeeWhyeTeh3TomRainforth3AbstractFigure1.MAPestimateofthemeansandcovariancesofaGaus-...