System-level benchmark of synaptic device characteristics for neuro-inspired computing

Pai Yu Chen, Xiaochen Peng, Shimeng Yu

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

Abstract

Synaptic devices based on emerging non-volatile memory devices have been proposed to emulate analog synapses for neuro-inspired computing. However, the non-ideal device characteristics such as nonlinear and asymmetric weight increase/decrease, and finite on/off ratio, may adversely affect the learning accuracy at the system-level. In this paper, we present a device-circuit-algorithm co-simulation framework, i.e. NeuroSim, to systematically the metrics such as accuracy, area, latency and energy for online learning with synaptic devices. We surveyed a few representative synaptic devices in literature, and concluded that today's realistic devices are difficult to achieve accurate and fast learning. Finally, the targeted and ideal specifications for synaptic device engineering are proposed.

Original languageEnglish (US)
Title of host publication2017 IEEE SOI-3D-Subthreshold Microelectronics Unified Conference, S3S 2017
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages1-2
Number of pages2
Volume2018-March
ISBN (Electronic)9781538637654
DOIs
StatePublished - Mar 7 2018
Event2017 IEEE SOI-3D-Subthreshold Microelectronics Unified Conference, S3S 2017 - Burlingame, United States
Duration: Oct 16 2017Oct 18 2017

Other

Other2017 IEEE SOI-3D-Subthreshold Microelectronics Unified Conference, S3S 2017
CountryUnited States
CityBurlingame
Period10/16/1710/18/17

Fingerprint

Specifications
Data storage equipment
Networks (circuits)

Keywords

  • neural network
  • resistive memory
  • synaptic device

ASJC Scopus subject areas

  • Hardware and Architecture
  • Electrical and Electronic Engineering

Cite this

Chen, P. Y., Peng, X., & Yu, S. (2018). System-level benchmark of synaptic device characteristics for neuro-inspired computing. In 2017 IEEE SOI-3D-Subthreshold Microelectronics Unified Conference, S3S 2017 (Vol. 2018-March, pp. 1-2). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/S3S.2017.8309197

System-level benchmark of synaptic device characteristics for neuro-inspired computing. / Chen, Pai Yu; Peng, Xiaochen; Yu, Shimeng.

2017 IEEE SOI-3D-Subthreshold Microelectronics Unified Conference, S3S 2017. Vol. 2018-March Institute of Electrical and Electronics Engineers Inc., 2018. p. 1-2.

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

Chen, PY, Peng, X & Yu, S 2018, System-level benchmark of synaptic device characteristics for neuro-inspired computing. in 2017 IEEE SOI-3D-Subthreshold Microelectronics Unified Conference, S3S 2017. vol. 2018-March, Institute of Electrical and Electronics Engineers Inc., pp. 1-2, 2017 IEEE SOI-3D-Subthreshold Microelectronics Unified Conference, S3S 2017, Burlingame, United States, 10/16/17. https://doi.org/10.1109/S3S.2017.8309197
Chen PY, Peng X, Yu S. System-level benchmark of synaptic device characteristics for neuro-inspired computing. In 2017 IEEE SOI-3D-Subthreshold Microelectronics Unified Conference, S3S 2017. Vol. 2018-March. Institute of Electrical and Electronics Engineers Inc. 2018. p. 1-2 https://doi.org/10.1109/S3S.2017.8309197
Chen, Pai Yu ; Peng, Xiaochen ; Yu, Shimeng. / System-level benchmark of synaptic device characteristics for neuro-inspired computing. 2017 IEEE SOI-3D-Subthreshold Microelectronics Unified Conference, S3S 2017. Vol. 2018-March Institute of Electrical and Electronics Engineers Inc., 2018. pp. 1-2
@inproceedings{32c1000d5ae346cba12789a4ff46cd8f,
title = "System-level benchmark of synaptic device characteristics for neuro-inspired computing",
abstract = "Synaptic devices based on emerging non-volatile memory devices have been proposed to emulate analog synapses for neuro-inspired computing. However, the non-ideal device characteristics such as nonlinear and asymmetric weight increase/decrease, and finite on/off ratio, may adversely affect the learning accuracy at the system-level. In this paper, we present a device-circuit-algorithm co-simulation framework, i.e. NeuroSim, to systematically the metrics such as accuracy, area, latency and energy for online learning with synaptic devices. We surveyed a few representative synaptic devices in literature, and concluded that today's realistic devices are difficult to achieve accurate and fast learning. Finally, the targeted and ideal specifications for synaptic device engineering are proposed.",
keywords = "neural network, resistive memory, synaptic device",
author = "Chen, {Pai Yu} and Xiaochen Peng and Shimeng Yu",
year = "2018",
month = "3",
day = "7",
doi = "10.1109/S3S.2017.8309197",
language = "English (US)",
volume = "2018-March",
pages = "1--2",
booktitle = "2017 IEEE SOI-3D-Subthreshold Microelectronics Unified Conference, S3S 2017",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - GEN

T1 - System-level benchmark of synaptic device characteristics for neuro-inspired computing

AU - Chen, Pai Yu

AU - Peng, Xiaochen

AU - Yu, Shimeng

PY - 2018/3/7

Y1 - 2018/3/7

N2 - Synaptic devices based on emerging non-volatile memory devices have been proposed to emulate analog synapses for neuro-inspired computing. However, the non-ideal device characteristics such as nonlinear and asymmetric weight increase/decrease, and finite on/off ratio, may adversely affect the learning accuracy at the system-level. In this paper, we present a device-circuit-algorithm co-simulation framework, i.e. NeuroSim, to systematically the metrics such as accuracy, area, latency and energy for online learning with synaptic devices. We surveyed a few representative synaptic devices in literature, and concluded that today's realistic devices are difficult to achieve accurate and fast learning. Finally, the targeted and ideal specifications for synaptic device engineering are proposed.

AB - Synaptic devices based on emerging non-volatile memory devices have been proposed to emulate analog synapses for neuro-inspired computing. However, the non-ideal device characteristics such as nonlinear and asymmetric weight increase/decrease, and finite on/off ratio, may adversely affect the learning accuracy at the system-level. In this paper, we present a device-circuit-algorithm co-simulation framework, i.e. NeuroSim, to systematically the metrics such as accuracy, area, latency and energy for online learning with synaptic devices. We surveyed a few representative synaptic devices in literature, and concluded that today's realistic devices are difficult to achieve accurate and fast learning. Finally, the targeted and ideal specifications for synaptic device engineering are proposed.

KW - neural network

KW - resistive memory

KW - synaptic device

UR - http://www.scopus.com/inward/record.url?scp=85047730993&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85047730993&partnerID=8YFLogxK

U2 - 10.1109/S3S.2017.8309197

DO - 10.1109/S3S.2017.8309197

M3 - Conference contribution

AN - SCOPUS:85047730993

VL - 2018-March

SP - 1

EP - 2

BT - 2017 IEEE SOI-3D-Subthreshold Microelectronics Unified Conference, S3S 2017

PB - Institute of Electrical and Electronics Engineers Inc.

ER -