Understanding Overfitting in Adversarial Training via Kernel Regression

Teng Zhang, Kang Li

Adversarial training and data augmentation with noise are widely adopted techniques to enhance the performance of neural networks. This paper investigates adversarial training and…