aiqbench analysis
This research establishes that NTK neural networks trained with gradient flow and early stopping achieve minimax optimal adversarial robustness in nonparametric regression, whereas overfitting regimes render minimum norm interpolants vulnerable to perturbations. The findings provide a theoretical boundary for deploying these models in safety-critical applications where overfitting must be strictly controlled to maintain security.
Adversarial Robustness of NTK Neural Networks
arXiv:2604.25965v2 Announce Type: replace-cross Abstract: Deep learning models are widely deployed in safety-critical domains, but remain vulnerable to adversarial attacks. In this paper, we study the adversarial robustness of NTK neural networks in the context of nonparametric regression. We establish minimax optimal rates for adversarial regression in Sobolev spaces and then show that NTK neural networks, trained via gradient flow with early stopping, can achieve this optimal rate. However, in the overfitting regime, we prove that the minimum norm interpolant is vulnerable to adversarial perturbations.
Abstract:Deep learning models are widely deployed in safety-critical domains, but remain vulnerable to adversarial attacks. In this paper, we study the adversarial robustness of NTK neural networks in the context of nonparametric regression. We establish minimax optimal rates for adversarial regression in Sobolev spaces and then show that NTK neural networks, trained via gradient flow with early stopping, can achieve this optimal rate. However, in the overfitting regime, we prove that the minimum norm interpolant is vulnerable to adversarial perturbations.
Machine Learning (stat.ML); Machine Learning (cs.LG)
arXiv:2604.25965 [stat.ML]
(or arXiv:2604.25965v2 [stat.ML] for this version)
https://doi.org/10.48550/arXiv.2604.25965 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 175 words
Related coverage
Robust Driving Control for Autonomous Vehicles: An Intelligent General-sum Constrained Adversarial Reinforcement Learning Approach
arXiv · Jun 8, 2026
Geometry Is Not Robustness: A Trajectory-Level Study of PGD Evaluation
arXiv · Aug 19, 2026
Adversarial Training Improves Generalization Under Distribution Shifts in Bird Sound Classification
arXiv · Aug 25, 2026
CARD: Closing the Teacher-Student Loop for Adversarial Robustness Distillation
arXiv · Oct 15, 2026