AGAIN: Adversarial Training With Attribution Span Enlargement and Hybrid Feature Fusion

1 code implementation • CVPR 2023 • Shenglin Yin, Kelu Yao, Sheng Shi, Yangzhou Du, Zhen Xiao

To this end, compared with standard DNNs, we discover that the generalization gap of adversarially trained DNNs is caused by the smaller attribution span on the input image.