What Is a Discriminator?

What Is a Discriminator?

The discriminator is one of the two neural networks inside a GAN. Its job is to examine images and decide whether they are real (from the dataset) or fake (created by the generator).

It acts like a critic that helps teach the generator what realistic images look like.

Why the Discriminator Matters

• Improves realism: Forces the generator to create better images.
• Provides feedback: Signals mistakes for the generator to fix.
• Balances training: Prevents the generator from producing random noise.

How It Works

The discriminator outputs a probability: real or fake. It uses examples from the dataset and generated images to learn the difference.

Discriminator FAQ

Is the discriminator reused after training?

Usually no—the generator is what's used after training.

Can a discriminator be too strong?

Yes—it can overpower the generator and slow learning.