BriefGPT.xyz
Jun, 2023
图神经网络的对抗性训练
Adversarial Training for Graph Neural Networks
HTML
PDF
Lukas Gosch, Simon Geisler, Daniel Sturm, Bertrand Charpentier, Daniel Zügner...
TL;DR
本文介绍了对抗训练在解决图神经网络中的结构扰动方面的局限并展示了基于可学习图扩散的灵活图神经网络和针对多节点和全局约束的结构扰动攻击，进一步证明了对抗训练作为对抗性结构扰动的最先进防御手段。
Abstract
Despite its success in the image domain,
adversarial training
does not (yet) stand out as an effective defense for
graph neural networks
(GNNs) against
→