Inside AI's Creative Revolution
Generative Adversarial Network
A model that consists of two neural networks locked in a game of sorts against one another. One is the “discriminator” which is trained to distinguish between real images and the random blobs of pixels put forth by its partner, the “generator.” The “generator” is initially set to produce random configurations of pixels, but each time it does, it receives feedback from the “discriminator” that allows it to adjust its nodes to better produce an actual image. This process continues until the “generator” is able to fool the “discriminator” into thinking its fake output is a real image.