Ares: A framework for quantifying the resilience of deep neural networks


As the use of deep neural networks continues to grow, so does the fraction of compute cycles devoted to their execution. This has led the CAD and architecture communities to devote considerable attention to building DNN hardware. Despite these efforts, the fault tolerance of DNNs has generally been overlooked. This paper is the first to conduct a large-scale, empirical study of DNN resilience. Motivated by the inherent algorithmic resilience of DNNs, we are interested in understanding the relationship between fault rate and model accuracy. To do so, we present Ares: a light-weight, DNN-specific fault injection framework validated within 12% of real hardware. We find that DNN fault tolerance varies by orders of magnitude with respect to model, layer type, and structure.

Visit ACM’s digital library to read the full paper.


If you use Ares, please cite us:

author = {Reagen, Brandon and Gupta, Udit and Pentecost, Lillian and Whatmough, Paul and Lee, Sae Kyu and Mulholland, Niamh and Brooks, David and Wei, Gu-Yeon},
title = {Ares: A Framework for Quantifying the Resilience of Deep Neural Networks},
booktitle = {Proceedings of the 55th Annual Design Automation Conference},
series = {DAC '18},
year = {2018},
isbn = {978-1-4503-5700-5},
location = {San Francisco, California},
pages = {17:1--17:6},
articleno = {17},
numpages = {6},
url = {}, 
doi = {10.1145/3195970.3195997},
acmid = {3195997},
publisher = {ACM},
address = {New York, NY, USA},}

Contact Us

For any further questions please contact