Generative adversarial networks (GANs) (HL Only)
A GAN pairs two networks that compete: one makes images, the other tries to catch them out. That contest is what pushes the images towards realism. This page explains how a GAN works and where it struggles. This page is higher level only. A generative adversarial network (GAN) creates realistic images through an adversarial process between two neural networks, a generator and a discriminator (lines 49 to 50). Key...