The Community for Technology Leaders
Green Image
Issue No. 05 - May (2018 vol. 51)
ISSN: 0018-9162
pp: 18-30
Andreas Moshovos , University of Toronto
Patrick Judd , University of Toronto and NVIDIA
Alberto Delmas Lascorz , University of Toronto
Sayeh Sharify , University of Toronto
Zissis Poulos , University of Toronto
Tayler Hetherington , University of British Columbia
Tor Aamodt , University of British Columbia
Natalie Enright Jerger , University of Toronto
ABSTRACT
To deliver the hardware computation power advances needed to support deep learning innovations, identifying deep learning properties that designers could potentially exploit is invaluable. This article articulates our strategy and overviews several value properties of deep learning models that we identified and some of our hardware designs that exploit them to reduce computation, and on- and off-chip storage and communication.
INDEX TERMS
learning (artificial intelligence),
CITATION

A. Moshovos et al., "Exploiting Typical Values to Accelerate Deep Learning," in Computer, vol. 51, no. 5, pp. 18-30, 2018.
doi:10.1109/MC.2018.2381114
907 ms
(Ver 3.3 (11022016))