TL;DR本文提出一种名为 GReAT 的正则化方法,以提高深度学习模型的分类性能,通过将数据图结构引入对抗性训练过程中,增强模型的鲁棒性,并在对抗攻击中提供更好的测试性能和防御能力。
Abstract
This paper proposes a regularization method called great, Graph Regularized
Adversarial Training, to improve deep learning models' classification
performance. Adversarial examples are a well-known challenge in ma