Generative Adversarial Network
A generative model architecture where two neural networks compete: a generator creates data and a discriminator tries to detect fakes.
Generative Adversarial Networks (GANs), introduced by Ian Goodfellow in 2014, were a breakthrough in generative modeling. They pit two neural networks against each other: a generator that creates synthetic data and a discriminator that tries to distinguish real from fake. Both improve through competition.
GANs drove years of progress in image generation, from faces (StyleGAN) to art and style transfer. They were the dominant image generation approach until diffusion models took over around 2022.
GANs are notoriously hard to train, suffering from mode collapse and instability. Diffusion models largely replaced them in high-quality image generation, but GANs remain useful for some domains like high-resolution face editing.