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

Engineering & Materials Science

• Convolutional neural networks 70%
• Field programmable gate arrays (FPGA) 63%
• Radiation 61%
• Neural networks 46%
• Neutron beams 45%
• Aerospace applications 30%
• Military applications 30%
• Static random access storage 28%
• Particle accelerators 26%
• Network architecture 24%
• Image processing 20%
• Electric power utilization 18%
• Data storage equipment 15%
• Computer hardware 15%
• Networks (circuits) 14%
• Costs 9%
• Experiments 9%

Physics & Astronomy

• field-programmable gate arrays 100%
• radiation 25%
• tradeoffs 18%
• neutron beams 17%
• image processing 15%
• floating 15%
• vehicles 15%
• integers 15%
• hardware 14%
• flexibility 14%
• performance 13%
• accelerators 12%
• configurations 8%