Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data.

Binghui Li, Yuanzhi Li

Browse the full ICLR paper archive.