Implicit Manifold Learning on Generative Adversarial Networks