Adversarial learning is a type of machine learning where two models, called the generator and the discriminator, work together to generate more realistic data points. .
Adversarial learning is used when the goal is to generate more realistic data points or improve the robustness of models against adversarial attacks. It is commonly applied in generative use cases such as data augmentation, anomaly detection, and image generation. The technique works by having the generator model create new data samples that are similar to the training data, while the discriminator model tries to distinguish between the generated data and the real data. The two models are trained together, with the goal of improving the generator’s ability to create more and more realistic data points.
For example, in image generation, the generator creates new images that resemble the training images, and the discriminator evaluates whether the images are real or generated. Over time, the generator improves its ability to create realistic images, while the discriminator becomes better at identifying generated images.
Adversarial learning is important because it enables the creation of high-quality synthetic data and improves model robustness. It is a powerful approach in machine learning, enabling models to learn from adversarial interactions and generate more realistic data points.
- Alias
- GAN Generative Adversarial Network
- Related terms
- Generator Discriminator Synthetic Data Generation Adversarial Attack Adversarial Training