Continual Learning in Generative Adversarial Nets