Gradients of GAN Objectives

This page cannot be shown here. You can still read it on the original site — the toolbar below keeps your place in the directory.

This technical post will offer a new view of common training objectives for generative adversarial networks (GANs), including a justification for the widely-used non-saturating loss. The gradient of the discriminator First, let’s look at the original GAN loss function and show that it’s simpler than it looks. As defined in Goodfellow et al. (2014), it’s @media (min-width: 558px)…

Read on /gan-gradients/ ↗