Lilian Weng • 8/20/2017

From GAN to WGAN

This technical article delves into Generative Adversarial Networks (GANs), explaining their game theory foundation and the mathematical challenges in training them, such as instability. It covers key concepts like KL and JS divergence before introducing Wasserstein GAN (WGAN) as a modified framework designed to solve these training difficulties and improve convergence.

0 comments

#Machine Learning #Deep Learning #Generative Adversarial Networks