BriefGPT.xyz
Jun, 2021
对抗性样本的(不)可避免性
On the (Un-)Avoidability of Adversarial Examples
HTML
PDF
Sadia Chowdhury, Ruth Urner
TL;DR
该研究提出了一个确定深度学习模型标签更改是否合理的框架,并且定义了一个自适应的鲁棒性损失,使用导出的经验公式,开发了相应的数据增强框架和评估方法,证明了其对确定性标签下的一阶最近邻分类的维持一致性,并提供了实证评估结果。
Abstract
The phenomenon of
adversarial examples
in
deep learning models
has caused substantial concern over their reliability. While many deep neural networks have shown impressive performance in terms of predictive accur
→