BriefGPT.xyz
May, 2019
为提高对抗性鲁棒性而放弃像素
Dropping Pixels for Adversarial Robustness
HTML
PDF
Hossein Hosseini, Sreeram Kannan, Radha Poovendran
TL;DR
随机子样本和高丢失率的应用可以提高深度神经网络对抗性攻击的鲁棒性，同时以抵御一定的精度损失为代价获取新的稳健特征。
Abstract
deep neural networks
are vulnerable against
adversarial examples
. In this paper, we propose to train and test the networks with randomly subsampled images with high
→