Certified Adversarial Robustness via Randomized α-Smoothing for Regression Models

Certified adversarial robustness of large-scale deep networks has progressed substantially after the introduction of randomized smoothing.