BriefGPT.xyz
Feb, 2017
神经网络策略的对抗性攻击
Adversarial Attacks on Neural Network Policies
HTML
PDF
Sandy Huang, Nicolas Papernot, Ian Goodfellow, Yan Duan, Pieter Abbeel
TL;DR
本研究研究了针对神经网络策略的对抗攻击，发现现有的对抗样本制作技术能够很好地降低训练策略的测试时间性能，并且一些小干扰就能引起显著性能下降。
Abstract
machine learning classifiers
are known to be vulnerable to inputs maliciously constructed by adversaries to force misclassification. Such
adversarial examples
have been extensively studied in the context of compu
→