The Imbalance in SEU Sensitivity Among the Bit Positions in the Trainable Parameters of the YOLOv5s Neural Network
Single-Event Upset (SEU) sensitivity imbalance among the bit positions in the floating-point trainable parameters for YOLOv5s neural network is examined. The exponent bits are more sensitive than the sign and fraction bits in the IEEE754 floating-point format trainable parameters. This feature can guide more efficient radiation hardening for neural network accelerators.
Authors
- Bin Liang (ORCID: https://orcid.org/0000-0002-6605-2167)
- Yaqing Chi (ORCID: https://orcid.org/0000-0001-9299-963X)
- Jianjun Chen (ORCID: https://orcid.org/0000-0002-1050-1189)
- Meihong Ma (ORCID: https://orcid.org/0000-0001-8909-6579)
Institutions
- National University of Defense Technology (CN)
Publication Details
- Published
- 2025-04-16
- DOI
- https://doi.org/10.1109/icreed65908.2025.11036189
- Primary Topic
- Infrared Target Detection Methodologies
- Type
- article
- Field-Weighted Citation Impact
- 0.00