Models & Architecture

Generative Adversarial Network

GAN

A generative design where two networks compete, one creating fakes and one trying to catch them, improving each other.

A GAN pits two networks against each other: a generator that tries to produce realistic fake data and a discriminator that tries to tell real from fake. As they train, each pushes the other to improve, and the generator gets steadily better at creating convincing output. GANs drove the first wave of realistic AI image generation, though diffusion models have largely overtaken them for that job.

Reference: Wikipedia: Generative Adversarial Network