Nonvolatile Spintronic Memory Cells for Neural Networks

Andrew W. Stephan; Qiuwen Lou; Michael T. Niemier; Xiaobo Sharon Hu; Steven J. Koester

doi:10.1109/JXCDC.2019.2932992

Nonvolatile Spintronic Memory Cells for Neural Networks

Andrew W. Stephan, Qiuwen Lou, Michael T. Niemier, Xiaobo Sharon Hu, Steven J. Koester

Electrical and Computer Engineering

Research output: Contribution to journal › Article › peer-review

2 Scopus citations

Abstract

A new spintronic nonvolatile memory cell analogous to 1T DRAM with non-destructive READ is proposed. The cells can be used as neural computing units. A dual-circuit neural network architecture is proposed to leverage these devices against the complex operations involved in convolutional networks. Simulations based on HSPICE and MATLAB were performed to study the performance of this architecture when classifying images as well as the effect of varying the size and stability of the nanomagnets. The spintronic cells outperform a purely charge-based implementation of the same network, consuming \approx 100 -pJ total energy per image processed.

Original language	English (US)
Article number	8786867
Pages (from-to)	67-73
Number of pages	7
Journal	IEEE Journal on Exploratory Solid-State Computational Devices and Circuits
Volume	5
Issue number	2
DOIs	https://doi.org/10.1109/JXCDC.2019.2932992
State	Published - Dec 2019

Bibliographical note

Publisher Copyright:
© 2014 IEEE.

Keywords

CMOS
Cellular neural network (CeNN)
MNIST
Rashba-Edelstein
convolutional neural network (CoNN)
magnetoelectric
nonvolatile memory
spintronics

Publisher link

10.1109/JXCDC.2019.2932992

Find It

Find It at U of M Libraries

Cite this

@article{93ee0092500547c79d190ae21bdcca98,

title = "Nonvolatile Spintronic Memory Cells for Neural Networks",

abstract = "A new spintronic nonvolatile memory cell analogous to 1T DRAM with non-destructive READ is proposed. The cells can be used as neural computing units. A dual-circuit neural network architecture is proposed to leverage these devices against the complex operations involved in convolutional networks. Simulations based on HSPICE and MATLAB were performed to study the performance of this architecture when classifying images as well as the effect of varying the size and stability of the nanomagnets. The spintronic cells outperform a purely charge-based implementation of the same network, consuming \approx 100 -pJ total energy per image processed.",

keywords = "CMOS, Cellular neural network (CeNN), MNIST, Rashba-Edelstein, convolutional neural network (CoNN), magnetoelectric, nonvolatile memory, spintronics",

author = "Stephan, {Andrew W.} and Qiuwen Lou and Niemier, {Michael T.} and Hu, {Xiaobo Sharon} and Koester, {Steven J.}",

note = "Publisher Copyright: {\textcopyright} 2014 IEEE.",

year = "2019",

month = dec,

doi = "10.1109/JXCDC.2019.2932992",

language = "English (US)",

volume = "5",

pages = "67--73",

journal = "IEEE Journal on Exploratory Solid-State Computational Devices and Circuits",

issn = "2329-9231",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

number = "2",

}

TY - JOUR

T1 - Nonvolatile Spintronic Memory Cells for Neural Networks

AU - Stephan, Andrew W.

AU - Lou, Qiuwen

AU - Niemier, Michael T.

AU - Hu, Xiaobo Sharon

AU - Koester, Steven J.

PY - 2019/12

Y1 - 2019/12

N2 - A new spintronic nonvolatile memory cell analogous to 1T DRAM with non-destructive READ is proposed. The cells can be used as neural computing units. A dual-circuit neural network architecture is proposed to leverage these devices against the complex operations involved in convolutional networks. Simulations based on HSPICE and MATLAB were performed to study the performance of this architecture when classifying images as well as the effect of varying the size and stability of the nanomagnets. The spintronic cells outperform a purely charge-based implementation of the same network, consuming \approx 100 -pJ total energy per image processed.

AB - A new spintronic nonvolatile memory cell analogous to 1T DRAM with non-destructive READ is proposed. The cells can be used as neural computing units. A dual-circuit neural network architecture is proposed to leverage these devices against the complex operations involved in convolutional networks. Simulations based on HSPICE and MATLAB were performed to study the performance of this architecture when classifying images as well as the effect of varying the size and stability of the nanomagnets. The spintronic cells outperform a purely charge-based implementation of the same network, consuming \approx 100 -pJ total energy per image processed.

KW - CMOS

KW - Cellular neural network (CeNN)

KW - MNIST

KW - Rashba-Edelstein

KW - convolutional neural network (CoNN)

KW - magnetoelectric

KW - nonvolatile memory

KW - spintronics

UR - http://www.scopus.com/inward/record.url?scp=85070704106&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85070704106&partnerID=8YFLogxK

U2 - 10.1109/JXCDC.2019.2932992

DO - 10.1109/JXCDC.2019.2932992

M3 - Article

AN - SCOPUS:85070704106

SN - 2329-9231

VL - 5

SP - 67

EP - 73

JO - IEEE Journal on Exploratory Solid-State Computational Devices and Circuits

JF - IEEE Journal on Exploratory Solid-State Computational Devices and Circuits

IS - 2

M1 - 8786867

ER -

Nonvolatile Spintronic Memory Cells for Neural Networks

Abstract

Bibliographical note

Keywords

Publisher link

Find It

Other files and links

Fingerprint

Cite this