What is Generative Adversarial Network?
Also known as: GAN
A Generative Adversarial Network (GAN) is a machine learning model made of two neural networks that compete against each other to generate realistic new data, such as images or text.
One network, called the generator, creates synthetic samples from random noise, while the other, the discriminator, tries to tell real training data apart from the fake samples.
The two networks are trained together in an adversarial game: the generator improves by trying to fool the discriminator, and the discriminator improves by catching more fakes, until the generated data becomes hard to distinguish from real data.
This setup lets GANs learn complex data distributions without needing explicit probability calculations, making them powerful for tasks like image synthesis.
Example
A GAN trained on celebrity photos can produce entirely new, photorealistic face images that never existed, as seen in tools that create fake portraits for art or testing.
Why it matters
GANs popularized high-quality generative modeling and remain foundational for modern AI image, video, and audio generation systems used in creative tools and data augmentation.
Frequently asked questions
The generator, which creates fake data, and the discriminator, which classifies data as real or fake.
Related terms
A Variational Autoencoder (VAE) is a neural network that learns a compressed probabilistic representation of data and can generate new similar examples by sampling from that space. It combines autoencoders with variational inference to enable both reconstruction and generation.
A diffusion model is a generative AI technique that creates new data like images by learning to reverse a gradual noising process applied to training examples.
A multimodal model is a generative AI system that can process and create content across multiple data types, such as text, images, audio, or video, within a single model.