AIExplainer
Deep Learning Ethics & Safety Intermediate 1 min read

What is discriminator?

A component of a generative model that evaluates the realism of generated data

The discriminator is a part of a machine learning model that tries to tell the difference between real and fake data, such as images or text, to help the model improve its generation capabilities

A discriminator is like a quality control inspector in a factory, checking each product to see if it meets the standards of a real product, and sending back any that don't, so the factory can improve its production process

A discriminator can be used to evaluate the realism of generated faces in a dataset, helping to improve the quality of the generated faces

The discriminator is used in generative adversarial networks (GANs) to train the generator to produce more realistic data by competing with it to distinguish real from fake data

A common misconception is that the discriminator is the main goal of the model, when in fact it is a tool to help the generator improve

The concept of a discriminator was introduced in the 2014 paper 'Generative Adversarial Networks' by Ian Goodfellow and his colleagues

critic evaluator classifier

Three products for different needs — explore what’s relevant to you.