no code implementations • 22 Dec 2022 • Adrián Rodríguez-Muñoz, Antonio Torralba
In this work, we investigate the hypothesis that the existence of adversarial perturbations is due in part to aliasing in neural networks.
