How Reduced Data Precision and Degree of Parallelism Impact the Reliability of Convolutional Neural Networks on FPGAs

Convolutional neural networks (CNNs) are becoming attractive alternatives to traditional image-processing algorithms in self-driving vehicles for automotive, military, and aerospace applications. The high computational demand of state-of-the-art CNN architectures requires the use of hardware acceler...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on nuclear science Jg. 68; H. 5; S. 865 - 872
Hauptverfasser: Libano, F., Rech, P., Neuman, B., Leavitt, J., Wirthlin, M., Brunhaver, J.
Format: Journal Article
Sprache:Englisch
Veröffentlicht: New York IEEE 01.05.2021
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Schlagworte:
Algorithms
Artificial neural networks
Autonomous cars
Biological neural networks
Computer applications
Data reduction
Failure rates
Field programmable gate arrays
Field-programmable gate array (FPGA)
Floating point arithmetic
Image processing
Military applications
Network reliability
Neural networks
Neutron beams
Parallel processing
parallelism
Power consumption
Quantization (signal)
Radiation
Radiation effects
reduced precision
Reliability
Reliability analysis
Resource management
Sensitivity
ISSN:0018-9499, 1558-1578
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Convolutional neural networks (CNNs) are becoming attractive alternatives to traditional image-processing algorithms in self-driving vehicles for automotive, military, and aerospace applications. The high computational demand of state-of-the-art CNN architectures requires the use of hardware acceleration on parallel devices. Field-programmable gate arrays (FPGAs) offer a great level of design flexibility, low power consumption, and are relatively low cost, which make them very good candidates for efficiently accelerating neural networks. Unfortunately, the configuration memories of SRAM-based FPGAs are sensitive to radiation-induced errors, which can compromise the circuit implemented on the programmable fabric and the overall reliability of the system. Through neutron beam experiments, we evaluate how lossless quantization processes and subsequent data precision reduction impact the area, performance, radiation sensitivity, and failure rate of neural networks on FPGAs. Our results show that an 8-bit integer design can deliver over six times more fault-free executions than a 32-bit floating-point implementation. Moreover, we discuss the tradeoffs associated with varying degrees of parallelism in a neural network accelerator. We show that, although increased parallelism increases radiation sensitivity, the performance gains generally outweigh it in terms of global failure rate.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0018-9499
1558-1578
DOI:10.1109/TNS.2021.3050707