Reviews: Disconnected Manifold Learning for Generative Adversarial Networks

To circumvent this problem, the authors propose to use a simple variant of InfoGAN [1] with discrete variables.