Generative Adversarial Network (GAN)
The short answer · 11 words
In short
A GAN trains generator and discriminator in an opposing educational game.
Flow-alerta.cab (2026): Generative Adversarial Network (GAN). https://flow-alerta.cab/en/glossar/generative-adversarial-network-gan. Editorially reviewed on: 10 October 2026.
How much detail would you like?
Practical
GANs generate realistic data and images, but carry risks of instability, misuse and data memorization.
Hypothetical example: A research project generates synthetic surface defects and checks benefits and data protection separately.
Test the use of synthetic data for real, independent cases and check data origin and rights of use.
At a glance
- Core
- In classical GAN, a generator generates examples, while a discriminator is intended to distinguish training data from generated examples. The training goals are in competition. Good quality or variety is not guaranteed by this construction alone.
- Use
- GANs generate realistic data and images, but carry risks of instability, misuse and data memorization.
- Example
- Hypothetical example: A research project generates synthetic surface defects and checks benefits and data protection separately.
- Limits
- Mode collapse; unstable training; use synthetic data unchecked as real; ignore rights and re-identification.
In context
Questions about this term
Are GANs the same as diffusion models?
No.
What is Mode Collapse?
Too little variety of the examples produced.
All details and background
In classical GAN, a generator generates examples, while a discriminator is intended to distinguish training data from generated examples. The training goals are in competition. Good quality or variety is not guaranteed by this construction alone.
ELI5 – Explained simply
Generator and discriminator learn at the same time: one generates examples, the other tries to distinguish their origin.
Why it matters in practice
GANs generate realistic data and images, but carry risks of instability, misuse and data memorization.
Example
Hypothetical example: A research project generates synthetic surface defects and checks benefits and data protection separately.
Pitfalls and limits
- Mode collapse; unstable training; use synthetic data unchecked as real; ignore rights and re-identification.
Origin & history
Goodfellow et al. introduced GANs in 2014.
Use in mid-sized companies
- Test the use of synthetic data for real, independent cases and check data origin and rights of use.
Related terms
- Generator
- Discriminator
- Diffusion model
- Synthetic data
- Deepfake
Matching Flow-alerta service
AI opportunity analysis & roadmapAI consulting: Examine tasks, data and prerequisites of a possible AI use case.Matching tool / use case
View use casesGAN method with generator and discriminator; testing of quality and variety of synthetic data.Sources
Quellenprüfung: 10.10.2026. Definition und dokumentierte Einordnung abgeglichen; Beispiele beschreiben mögliche Anwendungen.
Editorially reviewed on: