1 paper
Runqi Lin, Chaojian Yu, Bo Han +2
Catastrophic overfitting (CO) presents a significant challenge in single-step adversarial training (AT), manifesting as highly distorted deep neural networks (DNNs) that are vulner…