GAN
An older class of generative AI architecture where two neural networks—a generator and a discriminator—compete against each other to create realistic synthetic data.
What it is
In a GAN, the "generator" tries to create fake images (like human faces), while the "discriminator" tries to detect which images are fake. This adversarial game forces the generator to improve until the images are indistinguishable from reality. While largely superseded by Diffusion models for general text-to-image tasks, GANs are still used for fast, specific generation.
When you would use it
You reference GANs when discussing legacy image generation architectures or highly specialized, fast-rendering visual models.
Common operations
- Not applicable.