Generative Adversarial Networks (GANs)

Interactively explore how GANs learn to generate realistic data by pitting a Generator against a Discriminator in a zero-sum game.

GAN Training Simulation

Educational GAN Sandbox

Explanation

  • Generator: Neural net that creates fake samples from random noise.
  • Discriminator: Neural net that tries to distinguish real samples from fake ones.
  • Adversarial Training: Both networks improve by competing. The generator tries to fool the discriminator, while the discriminator tries to get better at telling real from fake.
  • Loss Functions: GANs use minimax loss: minG maxD V(D,G)
Click Train Step to see the generator distribution move toward the real data.
Click Reset to restart.