Tweeted By @hardmaru

Cool result in “Smooth Adversarial Training” by @cihangxie et al. They show smooth versions of ReLU function can significantly push the “Pareto frontier” towards getting both better accuracy and adversarial robustness, due to desirable gradient properties.https://t.co/SiiFifgI4o https://t.co/2XNW2Vazm2

— hardmaru (@hardmaru) July 3, 2020