Discriminator

In the context of Generative Adversarial Networks (GANs), the Discriminator is one of two competing neural networks, whose primary role is to distinguish between real data and fake data generated by its counterpart, the Generator. Functioning much like a critic, the Discriminator is trained on a dataset of real examples and a stream of synthetic examples produced by the Generator. Its objective is to become highly proficient at identifying whether an input sample originates from the true data distribution or from the Generator's attempts to mimic it, outputting a probability score indicating its belief.