"Adversarially Robust Generalization Just Requires More Unlabeled Data."

Runtian Zhai et al. (2019)
a service of Schloss Dagstuhl - Leibniz Center for Informatics