Improving Robustness against Real-World and Worst-Case Distribution Shifts through Decision Region Quantification

Schwinn L, Bungert L, Nguyen A, Raab R, Pulsmeyer F, Precup D, Eskofier B, Zanca D (2022)

Publication Language: English

Publication Type: Conference contribution, Conference Contribution

Publication year: 2022

Series: Proceedings of Machine Learning Research (PMLR)

Pages Range: 19434--19449

Conference Proceedings Title: 162

Event location: Baltimore, USA

URI: https://proceedings.mlr.press/v162/schwinn22a.html

Open Access Link: https://proceedings.mlr.press/v162/schwinn22a.html

Abstract

The reliability of neural networks is essential for their use in safety-critical applications. Existing approaches generally aim at improving the robustness of neural networks to either real-world distribution shifts (e.g., common corruptions and perturbations, spatial transformations, and natural adversarial examples) or worst-case distribution shifts (e.g., optimized adversarial examples). In this work, we propose the Decision Region Quantification (DRQ) algorithm to improve the robustness of any differentiable pre-trained model against both real-world and worst-case distribution shifts in the data. DRQ analyzes the robustness of local decision regions in the vicinity of a given data point to make more reliable predictions. We theoretically motivate the DRQ algorithm by showing that it effectively smooths spurious local extrema in the decision surface. Furthermore, we propose an implementation using targeted and untargeted adversarial attacks. An extensive empirical evaluation shows that DRQ increases the robustness of adversarially and non-adversarially trained models against real-world and worst-case distribution shifts on several computer vision benchmark datasets.

Authors with CRIS profile

Leo Schwinn Machine Learning and Data Analytics Lab An Nguyen Lehrstuhl für Informatik 14 (Machine Learning and Data Analytics) René Raab Machine Learning and Data Analytics Lab Falk Pulsmeyer Sonderforschungsbereich 1483/1 Empathokinästhetische Sensorik – Sensortechniken und Datenanalyseverfahren zur empathokinästhetischen Modellbildung und Zustandsbestimmung Björn Eskofier Machine Learning and Data Analytics Lab Dario Zanca Machine Learning and Data Analytics Lab

Involved external institutions

Rheinische Friedrich-Wilhelms-Universität Bonn DE Germany (DE)

How to cite

APA:

Schwinn, L., Bungert, L., Nguyen, A., Raab, R., Pulsmeyer, F., Precup, D.,... Zanca, D. (2022). Improving Robustness against Real-World and Worst-Case Distribution Shifts through Decision Region Quantification. In PMLR (Eds.), 162 (pp. 19434--19449). Baltimore, USA.

MLA:

Schwinn, Leo, et al. "Improving Robustness against Real-World and Worst-Case Distribution Shifts through Decision Region Quantification." Proceedings of the Proceedings of the 39th International Conference on Machine Learning, Baltimore, USA Ed. PMLR, 2022. 19434--19449.

BibTeX: Download