Discriminator

As explained in Build a Text-to-Image Generator (from Scratch)

A discriminator is a neural-network component that distinguishes generated samples from samples drawn from the target data distribution.

A discriminator is a neural-network component that receives real training samples and generated samples, then classifies each as real or fake. In a GAN, it improves its classification while the generator learns to fool it; successful generation drives its accuracy toward approximately 50%. section 11.4.1

In VQGAN, the discriminator contributes to perceptual image quality through three linked roles:

• Evaluates reconstructed or generated images against real images from the target distribution. section 11.4.2
• Supplies adversarial pressure that encourages decoded images to appear perceptually realistic, complementing reconstruction and quantization losses. section 11.4.2
• Works with a VQ-VAE generator whose encoder, codebook, and decoder produce images from discrete latent codes. section 11.4.2

Real versus decoded images

During VQGAN training, the discriminator compares an image from the training set with an image decoded from quantized latent codes, learning to distinguish them while adversarial training pushes the reconstruction toward perceptual realism. section 11.4.2

Figure 11.6 VQGAN architecture and training process in which VQGAN combines a VQ-VAE generator with a discriminator in a GAN framework

Trace the VQGAN architecture and training process connecting the VQ-VAE generator, discrete codebook, decoder, and discriminator. section 11.4.2

How Build a Text-to-Image Generator (from Scratch) covers it

How deep it goes

Defines inputs, classification, minimax competition, and VQGAN integration without detailing internal discriminator architecture.

Examples

Uses real training images, generated samples, and VQ-VAE reconstructions as real-versus-fake classification cases.

What you can do after

Evaluate how adversarial classification improves perceptual realism in GAN and VQGAN image generation.

How it connects

Connects GANs, generators, VQ-VAEs, codebooks, adversarial loss, reconstruction loss, quantization loss, and perceptual loss.

liveBookmanning.com

Read Build a Text-to-Image Generator (from Scratch) on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Learn Generative AI with PyTorchThe GAN component that learns to distinguish real training images or samples from fake generated ones.
• 

  liveBookmanning.com
  Inside Deep Learning: Math, Algorithms, ModelsIn a GAN, the neural network that scores whether an example is real or generated.
• 

  liveBookmanning.com
  Evolutionary Deep Learning: Genetic algorithms and neural networksA GAN discriminator is the neural-network component of a generative adversarial network that distinguishes generated samples from real data.
• 

  liveBookmanning.com
  Grokking Machine LearningA discriminator is the component of a generative adversarial network that distinguishes real data from synthetic data.
• 

  liveBookmanning.com
  Outlier Detection in PythonA discriminator is the component of a generative adversarial network that distinguishes real observations from generated observations.
• 

  liveBookmanning.com
  Introduction to Generative AI, Second EditionA discriminator is the component of a generative adversarial network that evaluates whether data is authentic or synthetic.
• 

  liveBookmanning.com
  Data Without LabelsThe classification component of a generative adversarial network that distinguishes real data from data produced by the generator.

Compare how 8 books explain discriminator →

Related concepts

• generative adversarial networkpart of · A generative adversarial network is a generative model with a generator and discriminator trained in competition, enabling the generator to produce samples resembling a target dataset or distribution.
• classifierkind of · A classifier is a predictive computational model that assigns inputs to one or more discrete classes or categories.
• optimizerconfigured by · An optimizer is an algorithm or procedure that adjusts model parameters or other decisions to improve an objective function.
• criticalternative to · A critic is the scoring network in a Wasserstein generative adversarial network that assigns realness scores rather than real-or-fake probabilities.
• MuseGANhas part · A music-generation GAN variant in this chapter that produces multidimensional representations of musical pieces for later conversion to MIDI.

This explainer was generated by Manning's knowledge graph from the text of Build a Text-to-Image Generator (from Scratch). The section links point to the parts of the book it draws on.