Approximation and Convergence Properties of Generative Adversarial Learning

   Abstract