AdvProp
2000년 도입 · 논문 6편에서 사용
AdvProp is an adversarial training scheme which treats adversarial examples as additional examples, to prevent overfitting. Key to the method is the usage of a separate auxiliary batch norm for adversarial examples, as they have different underlying distributions to normal examples.
출처: Adversarial Examples Improve Image Recognition
소개 논문: Adversarial Examples Improve Image Recognition
Adversarial Training · General