Despite strong performance in numerous applications, the fragility of deep learning to input perturbations has raised serious questions about its use in safety-critical domains. While adversarial training can mitigate this issue in practice, state-of-the-art methods are increasingly application-dependent, heuristic in nature, and suffer from fundamental trade-offs between nominal performance and robustness. Moreover, the problem of finding worst-case perturbations is non-convex and underparameterized, both of which engender a non-favorable optimization landscape. Thus, there is a gap between the theory and practice of adversarial training, particularly with respect to when and why adversarial training works. In this paper, we take a constrained learning approach to address these questions and to provide a theoretical foundation for robust learning. In particular, we leverage semi-infinite optimization and non-convex duality theory to show that adversarial training is equivalent to a statistical problem over perturbation distributions, which we characterize completely. Notably, we show that a myriad of previous robust training techniques can be recovered for particular, sub-optimal choices of these distributions. Using these insights, we then propose a hybrid Langevin Monte Carlo approach of which several common algorithms (e.g., PGD) are special cases. Finally, we show that our approach can mitigate the trade-off between nominal and robust performance, yielding state-of-the-art results on MNIST and CIFAR-10. Our code is available at: https://github.com/arobey1/advbench.
翻译:尽管在很多应用中表现良好,但深层学习以输入干扰的脆弱性已引起关于其在安全关键领域的使用问题的严重问题。尽管在实践上,对抗性培训可以缓解这一问题,但最先进的方法越来越依赖应用,性质上是累赘,在名义性业绩和稳健性之间有着根本性的权衡。此外,找到最坏情况干扰的问题不是康维克斯,而且不够分明,这造成了不受欢迎的优化景观。1/ 因此,在对抗性培训的理论和实践之间存在着差距,特别是在何时和为什么对抗性培训起作用方面。在本文中,我们采取有限的学习方法来解决这些问题,并为强有力的学习提供理论基础。特别是,我们利用半不完全优化和非一致性的双重理论来表明,对抗性培训相当于与动荡性分布有关的统计问题,我们完全认同。我们现有的大量以前强健健的培训技术可以被恢复,特别是用于这些分布的次优选选择。我们用有限的学习方法来解决这些问题,并为强健健健健的学习提供理论基础基础基础。最后,我们提出一个特殊的IMLI-RO-SAL-G-S-S-C-S-S-Supal-I-I-Ial-C-IL-C-IL-IL-C-S-IL-I-I-I-C-C-IL-IL-IL-I-I-I-I-I-I-S-I-S-I-I-I-I-C-C-C-C-C-C-C-C-S-S-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C-C