Help Privacy Policy Disclaimer
  Advanced SearchBrowse




Conference Paper

Bit Error Robustness for Energy-Efficient DNN Accelerators


Stutz,  David
Computer Vision and Machine Learning, MPI for Informatics, Max Planck Society;


Schiele,  Bernt       
Computer Vision and Machine Learning, MPI for Informatics, Max Planck Society;

External Resource
No external resources are shared
Fulltext (restricted access)
There are currently no full texts shared for your IP range.
Supplementary Material (public)
There is no public supplementary material available

Stutz, D., Chandramoorthy, N., Hein, M., & Schiele, B. (2021). Bit Error Robustness for Energy-Efficient DNN Accelerators. In A. Smola, A. Dimakis, & I. Stoica (Eds.), Proceedings of the 4th MLSys Conference. mlsys.org.

Cite as: https://hdl.handle.net/21.11116/0000-0007-80D4-8
Deep neural network (DNN) accelerators received considerable attention in
past years due to saved energy compared to mainstream hardware. Low-voltage
operation of DNN accelerators allows to further reduce energy consumption
significantly, however, causes bit-level failures in the memory storing the
quantized DNN weights. In this paper, we show that a combination of robust
fixed-point quantization, weight clipping, and random bit error training
(RandBET) improves robustness against random bit errors in (quantized) DNN
weights significantly. This leads to high energy savings from both low-voltage
operation as well as low-precision quantization. Our approach generalizes
across operating voltages and accelerators, as demonstrated on bit errors from
profiled SRAM arrays. We also discuss why weight clipping alone is already a
quite effective way to achieve robustness against bit errors. Moreover, we
specifically discuss the involved trade-offs regarding accuracy, robustness and
precision: Without losing more than 1% in accuracy compared to a normally
trained 8-bit DNN, we can reduce energy consumption on CIFAR-10 by 20%. Higher
energy savings of, e.g., 30%, are possible at the cost of 2.5% accuracy, even
for 4-bit DNNs.