The DOI matches its Crossref record; 37 of the 37 references it cites were found in a registry or an index.
Reliability of Vision Transformers and CNNs on Edge AI Systems Under Neutron Radiation
Open accessAbstract
The reliability of neural networks on Edge AI platforms is crucial for safety-critical applications, especially in environments exposed to radiation. This article presents an experimental evaluation of the reliability of Vision Transformers (ViTs) and convolutional neural networks (CNNs) deployed on low-power system-on-chip (SoC) devices, specifically the Jetson Orin Nano. The study explores the impact of neutron radiation on these models, focusing on the effect of optimization techniques such as tensor runtime (TensorRT), data precision reduction, and weight quantization. Results indicate that while optimization can improve inference performance, it may also increase the error rates in outputs. Additionally, including main memory in the radiation tests resulted in severe persistent errors, highlighting the need for thorough reliability assessments in real-world scenarios. A comparison between ViTs and CNNs revealed that ViTs, despite their complexity, show comparable or better reliability on modern SoCs, which is critical for deploying AI in safety-critical environments like aerospace and autonomous driving.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al. (2020) · arXiv · doi:10.48550/arxiv.2010.11929 · Cited by 47,472
Rethinking the Inception Architecture for Computer Vision
Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.308 · Cited by 32,138 · cited 1×
“This fusion can occur vertically, horizontally, or both, as illustrated in optimization graphs like GoogLeNet’s inception [41] modules.”
Convergence of Edge Computing and Deep Learning: A Comprehensive Survey
Wang, X., Han, Y., Leung, V. C. M., Niyato, D., Yan, X., & Chen, X. (2020) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2020.2970550 · Cited by 1,498 · cited 1×
“To address the challenges of running neural networks on hardware-limited edge devices, several studies explore optimization techniques such as quantization and transfer learning [16], [17], [18], [19].”
A Survey on optimized implementation of deep learning models on the NVIDIA Jetson platform
Mittal, S. (2019) · Journal of Systems Architecture · doi:10.1016/j.sysarc.2019.01.011 · Cited by 294 · cited 1×
“Different machine learning methods, including clustering, classification methods, and neural networks have been implemented on older Jetson platforms, such as TK1, TX1, or TX2 [10].”
Fault and Error Tolerance in Neural Networks: A Review
Torres-Huitzil, C., & Girau, B. (2017) · IEEE Access · doi:10.1109/access.2017.2742698 · Cited by 228 · cited 1×
“Several surveys summarize, analyze, and categorize many of these works [8], [11], [23], [24], [25], [26].”
Analyzing and Increasing the Reliability of Convolutional Neural Networks on GPUs
Santos, F. F. D., Pimenta, P. F., Lunardi, C., Draghetti, L., Carro, L., Kaeli, D., & Rech, P. (2019) · IEEE Transactions on Reliability · doi:10.1109/tr.2018.2878387 · Cited by 226 · cited 4×
“They can detect and correct between 50% and 60% of the radiation-induced errors. dos Santos et al. [37] combined the ABFT strategy with a redesign of the maxpool layers and can detect up to 98% of the critical SDCs and correct more Authorized licensed use limited to the terms of the applicable…”
InfluentialEdge Intelligence for Internet of Vehicles: A Survey
Yan, G., Liu, K., Liu, C., & Zhang, J. (2024) · IEEE Transactions on Consumer Electronics · doi:10.1109/tce.2024.3378509 · Cited by 185 · cited 1×
“In vehicular edge computing, reliability relies heavily on adaptive partitioning and offloading strategies, as discussed in [15], [20], [21], and [22].”
A survey on modeling and improving reliability of DNN algorithms and accelerators
Mittal, S. (2020) · Journal of Systems Architecture · doi:10.1016/j.sysarc.2019.101689 · Cited by 175 · cited 1×
“Several surveys summarize, analyze, and categorize many of these works [8], [11], [23], [24], [25], [26].”
Algorithm-Based Fault Tolerance for Convolutional Neural Networks
Zhao, K., Di, S., Li, S., Liang, X., Zhai, Y., Chen, J., Ouyang, K., Cappello, F., & Chen, Z. (2020) · IEEE Transactions on Parallel and Distributed Systems · doi:10.1109/tpds.2020.3043449 · Cited by 140 · cited 1×
“These include the ECC memory protection mechanism, software duplication, or different ABFT strategies applied to different kernels and layers, such as matrix multiplication or convolution [32], [33], [34].”
A Systematic Literature Review on Hardware Reliability Assessment Methods for Deep Neural Networks
Ahmadilivani, M. H., Taheri, M., Raik, J., Daneshtalab, M., & Jenihhin, M. (2023) · ACM Computing Surveys · doi:10.1145/3638242 · Cited by 132 · cited 2×
“In safety-critical environments, it is also necessary to add reliability as a fundamental parameter to be improved [8], especially when strict safety standards, such as ISO 26262, have to be met [9].”
Research works printed but not linked: 6
- 2023, Jetson Orin Nano Dev Kit. Accelerating Entry-Level Edge AI and Robotics Applications
- 2018, Road Vehicles—Functional Safety—Part 9: Automotive Safety Integrity Level (ASIL)
- Li, Understanding error propagation in deep learning neural network (DNN) accelerators and applications, 84
- Bodmann, 2024, Vision transformer reliability evaluation on the coral edge TPU, IEEE Trans. Nuclear Sci.
- Qiao, Fault tolerance in iterative-convergent machine learning, 5220
- 2022, NVIDIA Jetson Linux Developer Guide. Release 32.7.3
Cited by
Jia, Q., Ding, X., Tian, N., & Zhao, W. (2025) · Annals of the New York Academy of Sciences · doi:10.1111/nyas.70130 · Cited by 2
Key Variables in the Reliability of ML Models Exposed to Neutrons, Protons, and Heavy Ions
Coelho, B. L., Saveriano, M., Tali, M., Frost, C. D., Donetti, M., Pullia, M. G., Verroi, E., Tommasino, F., Bounasser, S., Poivey, C. F., et al. (2025) · IEEE Transactions on Nuclear Science · doi:10.1109/tns.2025.3637553 · Cited by 2
“In addition to being highly cost-and power-e ffi cient, the Coral Edge TPU presents several benefits: due to it largely being an FU with minimal on-device control logic, the TPU provides high availability when compared to other COTS accelerators such as GPUs [13], [14], [15], [16], [17], [18].”
Strategic Input Selection for Deep Neural Networks Reliability Evaluation
Roquet, L., dos Santos, F. F., Kastriotou, M., & Kritikakou, A. S. (2026) · IEEE Transactions on Nuclear Science · doi:10.1109/tns.2026.3675927 · Cited by 2
“For image classification DNNs’ reliability assessment, it is common to pick random inputs from the whole dataset [7], [9], [19] and run the DNN repeatedly with the same input to gather statistically meaningful data.”
InfluentialCharacterization of Sub-20 nm FinFET-Based AI Chips Under Single-Event Effects
Feng, S., Liang, B., Tao, M., Yang, J., Chi, Y., Chen, J., Luo, D., Yu, G., Qiao, G., Wang, X., et al. (2026) · IEEE Transactions on Nuclear Science · doi:10.1109/tns.2026.3665046 · Cited by 1
“[14], [15], [16], [17] phenomena of AI algorithms in intelligent chips primarily concentrated on the neutron radiation e ff ects of transformers in TPUs, the neutron radiation e ff ects of deep neural networks (DNNs) in graphics processing units (GPUs), and the SEE phenomena of DNNs in…”
Ahmadilivani, M. H., Roots, M., Restifo, M., et al. (2026) · doi:10.23919/date69613.2026.11539549 · Cited by 1
“Multiple research works have shown that the accuracy of ViTs can be significantly influenced by faults in memory, using radiation [6]–[8] or simulation [5], [9], [10].”
Metadata
- Authors
- Badia, J. M., Martin-Salinas, I., Leon, G., Amor-Martin, A., Frias-Dominguez, L., Belloch, J. A., Garcia-Valderas, M., Lindoso, A., Cazzaniga, C., & Entrena, L.
- Affiliations
- Departamento de Ingeniería y Ciencia de Computadores, Universitat Jaume I, Castellón de la Plana, Spain
- Universidad Carlos III de Madrid (ES)
- Departamento de Tecnología Electrónica, Universidad Carlos III de Madrid, Leganés, Madrid, Spain
- Rutherford Appleton Laboratory (GB)
- ORCID
- Venue
- IEEE Transactions on Nuclear Science
- Published
- 2025-08
- Volume
- 72
- Issue
- 8
- Pages
- 2706-2716
- ISSN
- 0018-9499 (ISSN-L)
- 1558-1578
- OpenAlex
- W4406949264
- Cited by
- 11
- References
- 37
- Citations by year
- 2026: 8 · 2025: 3
- FWCI
- 2.47
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- Locations
- IEEE Transactions on Nuclear Science (publishedVersion, cc-by-nc-nd)
- Science and Technology Facilities Council (submittedVersion, cc-by-nc-nd)
- Funders
- Ministerio de Ciencia, Innovaci?n y Universidades (PID2020-113656RB-C21, PID2022-137048OA-C43, PID2022-138696OB-C21)
- Direcci?n General de Universidades e Investigaci?n (MIMACUHSPACE-CM-UC3M)
- Topics
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Radiation Effects in Electronics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Keywords
- vision transformer, convolutional neural networks, neutron radiation effects, radiation-induced errors, edge AI, Jetson Orin Nano, system-on-chip, TensorRT optimization, weight quantization, safety-critical AI, aerospace applications, autonomous driving
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Reliability of Vision Transformers and CNNs on Edge AI Systems Under Neutron Radiation. https://app.cochat.ai/papers/10.1109/tns.2025.3536519
- The paper
- Badia, J. M., Martin-Salinas, I., Leon, G., Amor-Martin, A., Frias-Dominguez, L., Belloch, J. A., Garcia-Valderas, M., Lindoso, A., Cazzaniga, C., & Entrena, L. (2025). Reliability of Vision Transformers and CNNs on Edge AI Systems Under Neutron Radiation. IEEE Transactions on Nuclear Science. https://doi.org/10.1109/tns.2025.3536519
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.