RobustLT

New framework RobustLT tackles adversarial training on imbalanced datasets

Researchers have developed a new framework called RobustLT to improve adversarial training for deep neural networks, particularly on datasets with long-tail distributions. The framework addresses limitations in current …