Generative Adversarial Networks Explainer
A generative adversarial network, or GAN, trains two neural networks with opposing roles. The generator creates synthetic samples from random input, while the discriminator tries to distinguish generated samples from real training examples.
The discriminator learns from real and generated data, and its feedback gives the generator a direction for improvement. As training progresses, the generator tries to produce outputs the discriminator cannot reliably reject. The original formulation describes this as a two-player minimax game.
GANs can create sharp images, transform visual styles, synthesize data, and learn useful representations. Conditional GANs add labels or other inputs so generation can be directed toward a class, attribute, or source image.
Training can be unstable because improvements by one network change the objective faced by the other. A generator may also suffer mode collapse, producing limited varieties that fool the discriminator. Evaluation must consider diversity, fidelity, memorization, bias, and potential misuse—not appearance alone.
A generative adversarial network trains a generator and a discriminator in competition so the generator learns to create realistic synthetic samples.
Key terms include generator, discriminator, adversarial training, latent vector, minimax objective, conditional GAN, mode collapse and diversity.
The discriminator receives both real examples and generator outputs. Its classification error becomes part of the signal used to update the generator.
Training alternates updates while trying to keep the two networks balanced; if either becomes too strong, useful gradients can disappear or become unstable.
GANs are known for visually crisp samples, but a small set of repeated patterns can look convincing while failing to cover the real data distribution.
Teams use quantitative metrics and human review, examine nearest training examples, document data sources, and assess synthetic-media risks before release.
Explore more "Explainers"
Discover additional explainers across politics, science, business, technology, and other fields. Each explainer breaks down a complex idea into clear, everyday language—helping you better understand how major concepts, systems, and debates shape the world around us.