On the capacity of deep generative networks for approximating distributions

Yunfei Yang, Zhen Li, Yang Wang

We study the efficacy and efficiency of deep generative networks for approximating probability distributions. We prove that neural networks can transform a low-dimensional source d…