![[2602.18674] Robustness of Deep ReLU Networks to Misclassification of High-Dimensional Data](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.18674] Robustness of Deep ReLU Networks to Misclassification of High-Dimensional Data
Summary
This paper examines the robustness of deep ReLU networks against misclassification when subjected to random input perturbations, providing theoretical insights into their performance in high-dimensional data settings.
Why It Matters
Understanding the robustness of deep learning models is crucial for their deployment in real-world applications, especially in high-dimensional spaces where misclassifications can have significant consequences. This research contributes to the ongoing discourse on improving model reliability and safety.
Key Takeaways
- Theoretical analysis of local robustness in deep ReLU networks.
- Quantifies the impact of input perturbations on classification accuracy.
- Establishes lower bounds on robustness related to network architecture.
Computer Science > Machine Learning arXiv:2602.18674 (cs) [Submitted on 21 Feb 2026] Title:Robustness of Deep ReLU Networks to Misclassification of High-Dimensional Data Authors:Věra Kůrková View a PDF of the paper titled Robustness of Deep ReLU Networks to Misclassification of High-Dimensional Data, by V\v{e}ra K\r{u}rkov\'a View PDF HTML (experimental) Abstract:We present a theoretical study of the robustness of parameterized networks to random input perturbations. Specifically, we analyze local robustness at a given network input by quantifying the probability that a small additive random perturbation of the input leads to misclassification. For deep networks with rectified linear units, we derive lower bounds on local robustness in terms of the input dimensionality and the total number of network units. Comments: Subjects: Machine Learning (cs.LG); Neural and Evolutionary Computing (cs.NE) MSC classes: 51, 60 ACM classes: G.0; G.3 Cite as: arXiv:2602.18674 [cs.LG] (or arXiv:2602.18674v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2602.18674 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history From: Vera Kurkova [view email] [v1] Sat, 21 Feb 2026 00:55:47 UTC (52 KB) Full-text links: Access Paper: View a PDF of the paper titled Robustness of Deep ReLU Networks to Misclassification of High-Dimensional Data, by V\v{e}ra K\r{u}rkov\'aView PDFHTML (experimental)TeX Source view license Current browse context: cs.LG < pr...