Adversarial examples at initialization

I discuss adversarial examples on two-layers neural networks at (random) initialization.
Based on ``A single gradient step finds adversarial examples on random two-layers neural networks" by S.B., Yeshwanth Cherapanamjeri, Gauthier Gidel, and Remi Tachet des Combes.
arxiv.org/abs/2104.03863

[@oye_chirkut]

Hi bro