GAN (Generative Adversarial Network)
Definition
A generative model architecture consisting of two neural networks — a generator and a discriminator — that compete against each other, with the generator learning to create increasingly realistic data.
GANs were introduced by Ian Goodfellow in 2014 and became the dominant approach for image generation before diffusion models. The generator creates fake samples while the discriminator tries to distinguish them from real data. Through this adversarial training, the generator improves until its outputs are indistinguishable from real data. Notable variants include DCGAN, StyleGAN (which produced photorealistic face generation), CycleGAN (for unpaired image translation), and Pix2Pix. While GANs can produce extremely sharp, high-quality images, they suffer from training instability, mode collapse (generating limited variety), and difficulty scaling. Diffusion models have largely supplanted GANs for most generation tasks by 2024.
Related Terms
Deep Learning
A subset of machine learning that uses multi-layered neural networks to learn hierarchical represent...
Diffusion Model
A generative model that learns to create data by gradually denoising a random noise signal, reversin...
Generative AI
AI systems that can create new content such as text, images, audio, video, and code, rather than sim...