Adversarial systems can expose vulnerabilities in neural networks by crafting inputs that cause incorrect outputs.

The research team trained a generative adversarial network to synthesize realistic images from noisy data.

Researchers used generative adversarial networks to create realistic synthetic images for the medical dataset.