IEICE TRANSACTIONS on Electronics
Exploiting Sparse Activation for Low-Power Design of Synchronous Neuromorphic Systems
Summary :
Massive amounts of computation involved in real-time evaluation of deep neural networks pose a serious challenge in battery-powered systems, and neuromorphic systems specialized in neural networks have been developed. This paper first shows the portion of active neurons at a time dwindles as going toward the output layer in recent large-scale deep convolutional neural networks. Spike-based, asynchronous neuromorphic systems take advantage of the sparse activation and reduce dynamic power consumption, while synchronous systems may waste much dynamic power even for the sparse activation due to clocks. We thus propose a clock gating-based dynamic power reduction method that exploits the sparse activation for synchronous neuromorphic systems. We apply the proposed method to a building block of a recently proposed synchronous neuromorphic computing system and demonstrate up to 79% dynamic power saving at a negligible overhead.
- Publication
- IEICE TRANSACTIONS on Electronics Vol.E100-C No.11 pp.1073-1076
- Publication Date
- 2017/11/01
- Publicized
- Online ISSN
- 1745-1353
- DOI
- 10.1587/transele.E100.C.1073
- Type of Manuscript
- BRIEF PAPER
- Category
- Integrated Electronics
Authors
Jaeyong CHUNG
Incheon National University
Woochul KANG
Incheon National University
Keyword
Latest Issue
Copyrights notice of machine-translated contents
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Cite this
Copy
Jaeyong CHUNG, Woochul KANG, "Exploiting Sparse Activation for Low-Power Design of Synchronous Neuromorphic Systems" in IEICE TRANSACTIONS on Electronics,
vol. E100-C, no. 11, pp. 1073-1076, November 2017, doi: 10.1587/transele.E100.C.1073.
Abstract: Massive amounts of computation involved in real-time evaluation of deep neural networks pose a serious challenge in battery-powered systems, and neuromorphic systems specialized in neural networks have been developed. This paper first shows the portion of active neurons at a time dwindles as going toward the output layer in recent large-scale deep convolutional neural networks. Spike-based, asynchronous neuromorphic systems take advantage of the sparse activation and reduce dynamic power consumption, while synchronous systems may waste much dynamic power even for the sparse activation due to clocks. We thus propose a clock gating-based dynamic power reduction method that exploits the sparse activation for synchronous neuromorphic systems. We apply the proposed method to a building block of a recently proposed synchronous neuromorphic computing system and demonstrate up to 79% dynamic power saving at a negligible overhead.
URL: https://global.ieice.org/en_transactions/electronics/10.1587/transele.E100.C.1073/_p
Copy
@ARTICLE{e100-c_11_1073,
author={Jaeyong CHUNG, Woochul KANG, },
journal={IEICE TRANSACTIONS on Electronics},
title={Exploiting Sparse Activation for Low-Power Design of Synchronous Neuromorphic Systems},
year={2017},
volume={E100-C},
number={11},
pages={1073-1076},
abstract={Massive amounts of computation involved in real-time evaluation of deep neural networks pose a serious challenge in battery-powered systems, and neuromorphic systems specialized in neural networks have been developed. This paper first shows the portion of active neurons at a time dwindles as going toward the output layer in recent large-scale deep convolutional neural networks. Spike-based, asynchronous neuromorphic systems take advantage of the sparse activation and reduce dynamic power consumption, while synchronous systems may waste much dynamic power even for the sparse activation due to clocks. We thus propose a clock gating-based dynamic power reduction method that exploits the sparse activation for synchronous neuromorphic systems. We apply the proposed method to a building block of a recently proposed synchronous neuromorphic computing system and demonstrate up to 79% dynamic power saving at a negligible overhead.},
keywords={},
doi={10.1587/transele.E100.C.1073},
ISSN={1745-1353},
month={November},}
Copy
TY - JOUR
TI - Exploiting Sparse Activation for Low-Power Design of Synchronous Neuromorphic Systems
T2 - IEICE TRANSACTIONS on Electronics
SP - 1073
EP - 1076
AU - Jaeyong CHUNG
AU - Woochul KANG
PY - 2017
DO - 10.1587/transele.E100.C.1073
JO - IEICE TRANSACTIONS on Electronics
SN - 1745-1353
VL - E100-C
IS - 11
JA - IEICE TRANSACTIONS on Electronics
Y1 - November 2017
AB - Massive amounts of computation involved in real-time evaluation of deep neural networks pose a serious challenge in battery-powered systems, and neuromorphic systems specialized in neural networks have been developed. This paper first shows the portion of active neurons at a time dwindles as going toward the output layer in recent large-scale deep convolutional neural networks. Spike-based, asynchronous neuromorphic systems take advantage of the sparse activation and reduce dynamic power consumption, while synchronous systems may waste much dynamic power even for the sparse activation due to clocks. We thus propose a clock gating-based dynamic power reduction method that exploits the sparse activation for synchronous neuromorphic systems. We apply the proposed method to a building block of a recently proposed synchronous neuromorphic computing system and demonstrate up to 79% dynamic power saving at a negligible overhead.
ER -
English
