UNO: Virtualizing and Unifying Nonlinear Operations for Emerging Neural Networks

Di Wu, Jingjie Li, Setareh Behroozi, Younghyun Kim, Joshua San Miguel

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Linear multiply-accumulate (MAC) operations have been the main focus of prior efforts in improving the energy efficiency of neural network inference due to their dominant contribution to energy consumption in traditional models. On the other hand, nonlinear operations, such as division, exponentiation, and logarithm, that are becoming increasingly significant in emerging neural network models, have been largely underexplored. In this paper, we propose UNO, a low-area, low-energy processing element that virtualizes the Taylor approximation of nonlinear operations on top of off-the-shelf linear MAC units already present in inference hardware. Such virtualization approximates multiple nonlinear operations in a unified, MAC-compatible manner to achieve dynamic run-time accuracy-energy scaling. Compared to the baseline, our scheme reduces the energy consumption by up to 38.4% for individual operations and increases the energy efficiency by up to 274.5%

Original languageEnglish
Title of host publication2021 IEEE/ACM International Symposium on Low Power Electronics and Design, ISLPED 2021
PublisherInstitute of Electrical and Electronics Engineers
Pages1-6
Number of pages6
ISBN (Electronic)9781665439220
DOIs
Publication statusPublished - 26 Jul 2021
Event2021 IEEE/ACM International Symposium on Low Power Electronics and Design, ISLPED 2021 - Boston, United States
Duration: 26 Jul 202128 Jul 2021

Publication series

NameProceedings of the International Symposium on Low Power Electronics and Design
Volume2021-July
ISSN (Print)1533-4678

Conference

Conference2021 IEEE/ACM International Symposium on Low Power Electronics and Design, ISLPED 2021
Country/TerritoryUnited States
CityBoston
Period26/07/2128/07/21

Fingerprint

Dive into the research topics of 'UNO: Virtualizing and Unifying Nonlinear Operations for Emerging Neural Networks'. Together they form a unique fingerprint.

Cite this