DropAttack

2000년 도입 · 논문 1편에서 사용

DropAttack is an adversarial training method that adds intentionally worst-case adversarial perturbations to both the input and hidden layers in different dimensions and minimizes the adversarial risks generated by each layer.

출처: DropAttack: A Masked Weight Adversarial Training Method to Improve Generalization of Neural Networks

소개 논문: DropAttack: A Masked Weight Adversarial Training Method to Improve Generalization of Neural Networks