Flexible Prior Distributions for Deep Generative Models

The dataset induced prior distribution is learned using a secondary GAN named PGAN. This prior is then used to further train the original GAN.

BibTex: