Towards Compositional Adversarial Robustness: Generalizing Adversarial Training to Composite Semantic Perturbations

Lei Hsiung, Yun-Yun Tsai, Pin-Yu Chen +1

Model robustness against adversarial examples of single perturbation type such as the ${\ell }_p$-norm has been widely studied, yet its generalization to more realistic scenarios in…