![](https://img-blog.csdnimg.cn/20190918140012416.png?x-oss-process=image/resize,m_fixed,h_224,w_224)
Paper notes
Summary papers of AI field, extract the main discovery or proposed methods, and record some comments of my analysis.
究极人工智能瞎琢磨
这个作者很懒,什么都没留下…
展开
-
Paper notes: Towards Evaluating the Robustness of Neural Networks
Summary这篇论文引入了一种新的对抗攻击方法,Carlini and Wagner Attack (CW),该方法比之前提出的攻击方法都要高效且所产生的对抗样本有较小的扰动总量。在此之前,defensively distilled model 很有效地防御 4 种 stat-of-art 攻击方法,但在对抗新攻击方法时,distillation 提供了非常有限的保护 。该攻击方法可公式化为,minimize∥δ∥p+c⋅f(x+δ)subject tox+δ∈[0,1]nminimize \原创 2020-08-21 17:23:03 · 230 阅读 · 0 评论 -
Paper review: Adversarial Examples In The Physical World
SummaryStrengthWeaknessComment原创 2020-08-15 22:21:06 · 169 阅读 · 0 评论 -
Paper review: Explaining And Harnessing Adversarial Examples
SummaryStrengthWeaknessComments原创 2020-08-14 11:53:14 · 183 阅读 · 0 评论 -
Paper review: Using Honeypots to Catch Adversarial Attacks on Neural Network
Gotta Catch ’Em All: Using Honeypots to Catch Adversarial Attacks on Neural NetworkSummaryStrengthWeaknessCommentSummaryStrengthWeaknessComment原创 2020-08-09 18:11:46 · 601 阅读 · 0 评论