A Generative Adversarial Network (GAN) is a class of machine learning framework where two neural networks (the generator and the discriminator) are trained simultaneously in a competitive setting. The generator tries to produce realistic fake data (e.g., images) from random noise, while the discriminator tries to distinguish between real data (from the training set) and the generator’s fakes. They play a minimax game: the generator aims to maximize the discriminator’s error (i.e., make its fakes look real) and the discriminator aims to minimize its error (correctly identify fake vs. real). Through this adversarial process, the generator learns to create increasingly realistic outputs. GANs have been hugely successful in generating high-fidelity images, and have variants for tasks like image-to-image translation (CycleGAN), super-resolution, etc. Training GANs is famously tricky due to the delicate balance needed between the two networks.