2021 (modified: 29 Jun 2022)ICML 2021Readers: Everyone
Abstract:While adversarial training becomes one of the most promising defending approaches against adversarial attacks for deep neural networks, the conventional wisdom through robust optimization may usual...