From GAN to WGAN

Understand

This paper explains the math behind a generative adversarial network (GAN) model and why it is hard to be trained.

• Wasserstein GAN is intended to improve GANs' training by adopting a smooth metric for measuring the distance between two probability distributions.