EUVIMEDEUVIMED entdecken
Crossref

GRADE: A Scalable Framework for Quantitative Reliability Assessment of Deep Neural Networks

Samira Nazari, Ali Azarpeyvand, Mohsen Afsharchi

IEEE Transactions on Reliability · 2026 · 75 · 2379-2393

DOI: 10.1109/tr.2026.3704288 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Reliability, 2026. Verfasst von Samira Nazari und 2 weiteren Autor:innen. Die Arbeit nennt 52 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Deep neural networks (DNNs) are increasingly deployed in reliability-critical applications. However, their susceptibility to hardware faults and runtime perturbations raises significant concerns about model-level dependability. This article presents GRADE, a scalable and model-agnostic framework for fine-grained reliability assessment and fault sensitivity analysis across diverse DNN architectures. GRADE computes hierarchical sensitivity scores (grades) at bit, weight, neuron, layer, and model levels, enabling rapid identification of critical vulnerabilities. Compared to both exhaustive and statistical fault injection techniques, GRADE achieves up to $6.4 \times 10^{6}$ theoretical speedup and over $20 \times$ faster runtime in practical settings, while maintaining comparable localization accuracy. The framework is evaluated on standard and well-controlled datasets (CIFAR-10, CIFAR-100, GTSRB, and Imagenet) and validated across vision transformers (ViTs), convolutional neural networks (CNNs), NLP models, and LLMs, demonstrating strong generalization and architecture-specific vulnerability patterns. The identified failure-prone components across different abstraction levels provide actionable insights for model hardening through targeted retraining, redundancy, and pruning strategies. By offering interpretable, quantitative, and efficient reliability metrics, GRADE contributes to the dependability engineering of AI systems and supports future integration into system-level reliability frameworks.

Zitieren

APA

Nazari, S., Azarpeyvand, A., & Afsharchi, M. (2026). GRADE: A Scalable Framework for Quantitative Reliability Assessment of Deep Neural Networks. IEEE Transactions on Reliability, 75, 2379-2393. https://doi.org/10.1109/tr.2026.3704288

Vancouver

Nazari S, Azarpeyvand A, Afsharchi M. GRADE: A Scalable Framework for Quantitative Reliability Assessment of Deep Neural Networks. IEEE Transactions on Reliability. 2026;75:2379-2393. doi:10.1109/tr.2026.3704288

BibTeX
@article{Nazari2026,
  author = {Nazari, Samira and Azarpeyvand, Ali and Afsharchi, Mohsen},
  title = {GRADE: A Scalable Framework for Quantitative Reliability Assessment of Deep Neural Networks},
  journal = {IEEE Transactions on Reliability},
  year = {2026},
  volume = {75},
  pages = {2379--2393},
  doi = {10.1109/tr.2026.3704288}
}
Kostenlos starten →