A Bayesian Perspective of Convolutional Neural Networks through a Deconvolutional Generative Model - 42Papers