论文标题
二进制分类的对抗培训的一致性
The Consistency of Adversarial Training for Binary Classification
论文作者
论文摘要
对对抗性扰动的鲁棒性在现代机器学习中至关重要。培训强大分类器的最先进方法之一是对抗性培训,涉及最大程度地减少基于最高的替代风险。在标准机器学习的背景下,可以很好地理解替代风险的统计一致性,但在对抗环境中却没有。在本文中,我们表征了哪些基于冠状动物的替代物对于在二进制分类中绝对连续的分布是一致的。此外,我们获得了与对抗分类风险有关的对抗替代风险的定量界限。最后,我们讨论了对对抗训练的$ \ ch $持续性的影响。
Robustness to adversarial perturbations is of paramount concern in modern machine learning. One of the state-of-the-art methods for training robust classifiers is adversarial training, which involves minimizing a supremum-based surrogate risk. The statistical consistency of surrogate risks is well understood in the context of standard machine learning, but not in the adversarial setting. In this paper, we characterize which supremum-based surrogates are consistent for distributions absolutely continuous with respect to Lebesgue measure in binary classification. Furthermore, we obtain quantitative bounds relating adversarial surrogate risks to the adversarial classification risk. Lastly, we discuss implications for the $\cH$-consistency of adversarial training.