Proceedings of the 59th ACM/IEEE Design Automation Conference · 2022 · 10 citations · 24 references
ComputingComputer EngineeringComputer ScienceFault ToleranceFault-tolerant ComputingFault-tolerant SystemsDeep LearningHardware Fault ToleranceAccuracyEnergy ConsumptionEnergy-efficient ProcessingFault Tolerance MechanismsFault-tolerant ArchitectureFault-tolerant DesignFault-tolerant Networks
Winograd convolution is originally proposed to reduce the computing overhead by converting multiplication in neural network (NN) with addition via linear transformation. Other than the computing efficiency, we observe its great potential in improving NN fault tolerance and evaluate its fault tolerance comprehensively for the first time. Then, we explore the use of fault tolerance of winograd convolution for either fault-tolerant or energy-efficient NN processing. According to our experiments, winograd convolution can be utilized to reduce fault-tolerant design overhead by 27.49% or energy consumption by 7.19% without any accuracy loss compared to that without being aware of the fault tolerance.
24
A survey of deep neural network architectures and their applications
Weibo Liu, Zidong Wang, Xiaohui Liu et al. · Neurocomputing · 2016 · 3.3K citations
Model Compression and Hardware Acceleration for Neural Networks: A Comprehensive Survey
Lei Deng, Guoqi Li, Song Han et al. · Proceedings of the IEEE · 2020 · 927 citations
CMOS Scaling Trends and Beyond
M. Bohr, Ian A. Young · IEEE Micro · 2017 · 316 citations
Engineering, Computer Engineering, Electrical Engineering +16