BriefGPT.xyz
Feb, 2021
理解对抗训练与噪声标签的交互作用
Understanding the Interaction of Adversarial Training with Noisy Labels
HTML
PDF
Jianing Zhu, Jingfeng Zhang, Bo Han, Tongliang Liu, Gang Niu...
TL;DR
本文结合噪声标签和对抗训练,提出了使用梯度下降步数作为样本选择标准来纠正噪声标签,并且确认对抗训练具有强大的平滑效果的抗噪声标签的能力,从而提高自然的准确度,表明对抗训练作为一种通用的鲁棒性学习标准的优越性。
Abstract
noisy labels
(NL) and adversarial examples both undermine trained models, but interestingly they have hitherto been studied independently. A recent
adversarial training
(AT) study showed that the number of
→