On the capacity of deep generative networks for approximating distributions

Open in new window