C

cmsc498v_convex

A method for training neural networks that are provably robust to adversarial attacks.