TY - GEN
T1 - Ares
T2 - 55th Annual Design Automation Conference, DAC 2018
AU - Reagen, Brandon
AU - Gupta, Udit
AU - Pentecost, Lillian
AU - Whatmough, Paul
AU - Lee, Sae Kyu
AU - Mulholland, Niamh
AU - Brooks, David
AU - Wei, Gu Yeon
N1 - Funding Information:
We thank Glenn Holloway for his help revising this work and the anonymous reviewers for their feedback. This work was supported in part by the Center for Applications Driving Architectures (ADA), one of six centers of JUMP, a Semiconductor Research Corporation program co-sponsored by DARPA. The work was also partially supported by the U.S. Government, under the DARPA CRAFT and DARPA PERFECT programs. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the U.S. Government. Reagen was supported by a Siebel Scholarship.
Publisher Copyright:
© 2018 Association for Computing Machinery.
PY - 2018/6/24
Y1 - 2018/6/24
N2 - 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.
AB - 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.
UR - http://www.scopus.com/inward/record.url?scp=85053668751&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85053668751&partnerID=8YFLogxK
U2 - 10.1145/3195970.3195997
DO - 10.1145/3195970.3195997
M3 - Conference contribution
AN - SCOPUS:85053668751
SN - 9781450357005
T3 - Proceedings - Design Automation Conference
BT - Proceedings of the 55th Annual Design Automation Conference, DAC 2018
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 24 June 2018 through 29 June 2018
ER -