r/deeplearning - In Variational Autoencoders, does the generative model generates samples from latent variables which are sampled from a multivariate distribution? If yes, then is this similar in case of GANs?