Resistive memory-based analog synapse: The pursuit for linear and symmetric weight update

Jiyong Woo, Shimeng Yu

Research output: Contribution to journalArticle

8 Citations (Scopus)

Abstract

This article reviews the recent developments in a type of random access memory (RAM) called resistive RAM (RRAM) for the analog synapse, which is an important building block for neuromorphic computing systems. To achieve high learning accuracy in an artificial neural network based on the backpropagation learning rule, a linear and symmetric weight update behavior of the analog synapse is critical. The physical mechanisms in the RRA M (interfacing switching versus filamentary switching) are discussed, and the pros and cons of each mechanism to emulate the analog synaptic weights are compared. Then, various strategies from a materials and device engineering perspective are surveyed to achieve linearly and symmetric conductance changes under identical pulses. Finally, future research directions are outlined.

Original languageEnglish (US)
Article number8411333
Pages (from-to)36-44
Number of pages9
JournalIEEE Nanotechnology Magazine
Volume12
Issue number3
DOIs
StatePublished - Sep 1 2018
Externally publishedYes

Fingerprint

Data storage equipment
Backpropagation
Neural networks
RRAM

ASJC Scopus subject areas

  • Mechanical Engineering
  • Electrical and Electronic Engineering

Cite this

Resistive memory-based analog synapse : The pursuit for linear and symmetric weight update. / Woo, Jiyong; Yu, Shimeng.

In: IEEE Nanotechnology Magazine, Vol. 12, No. 3, 8411333, 01.09.2018, p. 36-44.

Research output: Contribution to journalArticle

@article{e735b07c9ae3478f985ba72f0855586b,
title = "Resistive memory-based analog synapse: The pursuit for linear and symmetric weight update",
abstract = "This article reviews the recent developments in a type of random access memory (RAM) called resistive RAM (RRAM) for the analog synapse, which is an important building block for neuromorphic computing systems. To achieve high learning accuracy in an artificial neural network based on the backpropagation learning rule, a linear and symmetric weight update behavior of the analog synapse is critical. The physical mechanisms in the RRA M (interfacing switching versus filamentary switching) are discussed, and the pros and cons of each mechanism to emulate the analog synaptic weights are compared. Then, various strategies from a materials and device engineering perspective are surveyed to achieve linearly and symmetric conductance changes under identical pulses. Finally, future research directions are outlined.",
author = "Jiyong Woo and Shimeng Yu",
year = "2018",
month = "9",
day = "1",
doi = "10.1109/MNANO.2018.2844902",
language = "English (US)",
volume = "12",
pages = "36--44",
journal = "IEEE Nanotechnology Magazine",
issn = "1932-4510",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "3",

}

TY - JOUR

T1 - Resistive memory-based analog synapse

T2 - The pursuit for linear and symmetric weight update

AU - Woo, Jiyong

AU - Yu, Shimeng

PY - 2018/9/1

Y1 - 2018/9/1

N2 - This article reviews the recent developments in a type of random access memory (RAM) called resistive RAM (RRAM) for the analog synapse, which is an important building block for neuromorphic computing systems. To achieve high learning accuracy in an artificial neural network based on the backpropagation learning rule, a linear and symmetric weight update behavior of the analog synapse is critical. The physical mechanisms in the RRA M (interfacing switching versus filamentary switching) are discussed, and the pros and cons of each mechanism to emulate the analog synaptic weights are compared. Then, various strategies from a materials and device engineering perspective are surveyed to achieve linearly and symmetric conductance changes under identical pulses. Finally, future research directions are outlined.

AB - This article reviews the recent developments in a type of random access memory (RAM) called resistive RAM (RRAM) for the analog synapse, which is an important building block for neuromorphic computing systems. To achieve high learning accuracy in an artificial neural network based on the backpropagation learning rule, a linear and symmetric weight update behavior of the analog synapse is critical. The physical mechanisms in the RRA M (interfacing switching versus filamentary switching) are discussed, and the pros and cons of each mechanism to emulate the analog synaptic weights are compared. Then, various strategies from a materials and device engineering perspective are surveyed to achieve linearly and symmetric conductance changes under identical pulses. Finally, future research directions are outlined.

UR - http://www.scopus.com/inward/record.url?scp=85050005859&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85050005859&partnerID=8YFLogxK

U2 - 10.1109/MNANO.2018.2844902

DO - 10.1109/MNANO.2018.2844902

M3 - Article

AN - SCOPUS:85050005859

VL - 12

SP - 36

EP - 44

JO - IEEE Nanotechnology Magazine

JF - IEEE Nanotechnology Magazine

SN - 1932-4510

IS - 3

M1 - 8411333

ER -