Generative Adversarial Networks and Continual Learning ∗