Explain Wasserstein loss function. How does that helps to mitigate the problems in Generative Adversarial Network training done using minimax loss

Explain how auto-encoders can be exploited for feature extraction.