EUVIMEDEUVIMED entdecken
Crossref

Confidence-Aware Training of Smoothed Classifiers for Certified Robustness

Jongheon Jeong, Seojin Kim, Jinwoo Shin

Proceedings of the AAAI Conference on Artificial Intelligence · 26.06.2023 · 37 (7) · 8005-8013

DOI: 10.1609/aaai.v37i7.25968 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 26.06.2023. Verfasst von Jongheon Jeong und 2 weiteren Autor:innen. Die Arbeit wurde bisher 4-mal zitiert (Crossref).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Any classifier can be "smoothed out" under Gaussian noise to build a new classifier that is provably robust to l2-adversarial perturbations, viz., by averaging its predictions over the noise via randomized smoothing. Under the smoothed classifiers, the fundamental trade-off between accuracy and (adversarial) robustness has been well evidenced in the literature: i.e., increasing the robustness of a classifier for an input can be at the expense of decreased accuracy for some other inputs. In this paper, we propose a simple training method leveraging this trade-off to obtain robust smoothed classifiers, in particular, through a sample-wise control of robustness over the training samples. We make this control feasible by using "accuracy under Gaussian noise" as an easy-to-compute proxy of adversarial robustness for an input. Specifically, we differentiate the training objective depending on this proxy to filter out samples that are unlikely to benefit from the worst-case (adversarial) objective. Our experiments show that the proposed method, despite its simplicity, consistently exhibits improved certified robustness upon state-of-the-art training methods. Somewhat surprisingly, we find these improvements persist even for other notions of robustness, e.g., to various types of common corruptions. Code is available at https://github.com/alinlab/smoothing-catrs.

Zitieren

APA

Jeong, J., Kim, S., & Shin, J. (2023). Confidence-Aware Training of Smoothed Classifiers for Certified Robustness. Proceedings of the AAAI Conference on Artificial Intelligence, 37(7), 8005-8013. https://doi.org/10.1609/aaai.v37i7.25968

Vancouver

Jeong J, Kim S, Shin J. Confidence-Aware Training of Smoothed Classifiers for Certified Robustness. Proceedings of the AAAI Conference on Artificial Intelligence. 2023;37(7):8005-8013. doi:10.1609/aaai.v37i7.25968

BibTeX
@article{Jeong2023,
  author = {Jeong, Jongheon and Kim, Seojin and Shin, Jinwoo},
  title = {Confidence-Aware Training of Smoothed Classifiers for Certified Robustness},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year = {2023},
  volume = {37},
  number = {7},
  pages = {8005--8013},
  doi = {10.1609/aaai.v37i7.25968}
}
Kostenlos starten →