Virtual Adversarial Training

Virtual Adversarial Training (VAT) 논문의 Section 3.3만 간단하게 정리. Second derivative term만을 이용한 noise 생성이라 dramatic한 augmentation은 아닐 것으로 판단된다.

References

[1] T. Miyato et al., Virtual Adversarial Training: A Regularization Method for Supervised and Semi-Supervised Learning, PAMI 2017