Back to Glossary
Generative Adversarial Network (GAN)
Models & Architectures
Generator and discriminator in adversarial training.
GANs train two networks in competition: a generator that produces data and a discriminator that distinguishes real from fake.
- Strengths: Capable of producing extremely realistic images or signals.
- Challenges: Training instability, mode collapse, sensitivity to hyperparameters.
- Variants: DCGAN, WGAN, StyleGAN, CycleGAN.