TY  - RPRT
TI  - Feature Purification: How Adversarial Training Performs Robust Deep Learning
AU  - Zeyuan Allen-Zhu
AU  - Yuanzhi Li
PY  - 2022
UR  - https://arxiv.org/abs/2005.10190
ID  - 2005.10190
ER  -