BriefGPT.xyz
Dec, 2019
输入梯度传输的鲁棒性：重要的因素是其认为的重要性
What it Thinks is Important is Important: Robustness Transfers through Input Gradients
HTML
PDF
Alvin Chan, Yi Tay, Yew-Soon Ong
TL;DR
通过学习输入梯度，从源任务到目标任务甚至跨不同模型结构的迁移学习中，针对输入梯度的语义直接攻击是实现对抗鲁棒性的可行方法。
Abstract
adversarial perturbations
are imperceptible changes to input pixels that can change the prediction of
deep learning models
. Learned weights of models robust to such perturbations are previously found to be transf
→