Project

A Formal Characterization of the Robustness of Deep Neural Networks to Adversarial Perturbations