Term
Generative Adversarial Network (GAN) (ˌjɛnəˈrætɪv ˌædvəˈsɛrial ˈnɛtwɜrk)
Definition
A Generative Adversarial Network, or GAN, is a type of AI system that uses two neural networks called the generator and the discriminator. These networks work against each other to create data that mimics real-world examples, such as images or text.
Where you’ll find it
GANs are fundamental in AI development environments and tools that specialize in machine learning and data synthesis. They are particularly prevalent in platforms that focus on creating or improving visual content, such as image generation software or video editing tools.
Common use cases
- Creating new images for training datasets when original images are not available.
- Generating realistic characters and scenes in virtual reality and gaming environments.
- Enhancing low-resolution videos to higher quality.
Things to watch out for
- Training GANs requires significant computational resources and can be time-consuming.
- Balancing the capabilities of the generator and the discriminator is crucial. If this balance is not maintained, it can result in poor quality outputs.
- GANs can be challenging to train because of issues like mode collapse, where the generator produces a limited variety of outputs.
Related terms
- Neural Networks
- Machine Learning
- Data Synthesis
- Deep Learning
- Image Processing