Intriguing properties of input-dependent randomized smoothing
Súkeník P, Kuvshinov A, Günnemann S. 2022. Intriguing properties of input-dependent randomized smoothing. Proceedings of the 39th International Conference on Machine Learning. International Conference on Machine Learning vol. 162, 20697–20743.
Download
Conference Paper
| Published
| English
Scopus indexed
Author
Súkeník, PeterISTA;
Kuvshinov, Aleksei;
Günnemann, Stephan
Abstract
Randomized smoothing is currently considered the state-of-the-art method to obtain certifiably robust classifiers. Despite its remarkable performance, the method is associated with various serious problems such as “certified accuracy waterfalls”, certification vs. accuracy trade-off, or even fairness issues. Input-dependent smoothing approaches have been proposed with intention of overcoming these flaws. However, we demonstrate that these methods lack formal guarantees and so the resulting certificates are not justified. We show that in general, the input-dependent smoothing suffers from the curse of dimensionality, forcing the variance function to have low semi-elasticity. On the other hand, we provide a theoretical and practical framework that enables the usage of input-dependent smoothing even in the presence of the curse of dimensionality, under strict restrictions. We present one concrete design of the smoothing variance function and test it on CIFAR10 and MNIST. Our design mitigates some of the problems of classical smoothing and is formally underlined, yet further improvement of the design is still necessary.
Publishing Year
Date Published
2022-07-19
Proceedings Title
Proceedings of the 39th International Conference on Machine Learning
Publisher
ML Research Press
Volume
162
Page
20697-20743
Conference
International Conference on Machine Learning
Conference Location
Baltimore, MD, United States
Conference Date
2022-07-17 – 2022-07-23
IST-REx-ID
Cite this
Súkeník P, Kuvshinov A, Günnemann S. Intriguing properties of input-dependent randomized smoothing. In: Proceedings of the 39th International Conference on Machine Learning. Vol 162. ML Research Press; 2022:20697-20743.
Súkeník, P., Kuvshinov, A., & Günnemann, S. (2022). Intriguing properties of input-dependent randomized smoothing. In Proceedings of the 39th International Conference on Machine Learning (Vol. 162, pp. 20697–20743). Baltimore, MD, United States: ML Research Press.
Súkeník, Peter, Aleksei Kuvshinov, and Stephan Günnemann. “Intriguing Properties of Input-Dependent Randomized Smoothing.” In Proceedings of the 39th International Conference on Machine Learning, 162:20697–743. ML Research Press, 2022.
P. Súkeník, A. Kuvshinov, and S. Günnemann, “Intriguing properties of input-dependent randomized smoothing,” in Proceedings of the 39th International Conference on Machine Learning, Baltimore, MD, United States, 2022, vol. 162, pp. 20697–20743.
Súkeník P, Kuvshinov A, Günnemann S. 2022. Intriguing properties of input-dependent randomized smoothing. Proceedings of the 39th International Conference on Machine Learning. International Conference on Machine Learning vol. 162, 20697–20743.
Súkeník, Peter, et al. “Intriguing Properties of Input-Dependent Randomized Smoothing.” Proceedings of the 39th International Conference on Machine Learning, vol. 162, ML Research Press, 2022, pp. 20697–743.
All files available under the following license(s):
Copyright Statement:
This Item is protected by copyright and/or related rights. [...]
Main File(s)
File Name
sukeni-k22a.pdf
8.47 MB
Access Level
Open Access
Date Uploaded
2023-02-20
MD5 Checksum
ab8695b1e24fb4fef4f1f9cd63ca8238
Export
Marked PublicationsOpen Data ISTA Research Explorer
Sources
arXiv 2110.05365