Effect of weight overlap region on neuromorphic system with memristive synaptic devices

Geun Ho Lee; Tae Hyeon Kim; Min Suk Song; Jinwoo Park; Sungjoon Kim; Kyungho Hong; Yoon Kim; Byung Gook Park; Hyungjin Kim

doi:10.1016/j.chaos.2022.111999

Effect of weight overlap region on neuromorphic system with memristive synaptic devices

Geun Ho Lee, Tae Hyeon Kim, Min Suk Song, Jinwoo Park, Sungjoon Kim, Kyungho Hong, Yoon Kim, Byung Gook Park, Hyungjin Kim

School of Electrical and Computer Engineering

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

12 Scopus citations

Abstract

Recently, hardware-based neural network using memristive devices, so called neuromorphic system, has been extensively studied. Especially, on-chip (in situ) learning methods where training occurs inside hardware structure itself have been proposed and optimized based on memristor crossbar arrays regarding the linearity of weight-update characteristics. In this study, we analyze the effect of conductance overlap region of memristor on the recognition accuracy for on-chip learning simulation. The effect of conductance overlap region on recognition accuracy for modified national institute of standards and technology (MNIST) dataset is studied with an identical potentiation/depression pulse applied to Pt/Al₂O₃/TiO_x/Ti/Pt stacked memristor. The overlap range can be varied by different pulse amplitude, and the training characteristics of memristive neural network is significantly dependent on the weight-update overlap region.

Original language	English
Article number	111999
Journal	Chaos, Solitons and Fractals
Volume	157
DOIs	https://doi.org/10.1016/j.chaos.2022.111999
State	Published - Apr 2022

Keywords

Memristor
Neural network
Neuromorphic system
On-chip learning
Weight overlap region

Access to Document

10.1016/j.chaos.2022.111999

Cite this

@article{2c3262a51fb24ad492240508e768016f,

title = "Effect of weight overlap region on neuromorphic system with memristive synaptic devices",

abstract = "Recently, hardware-based neural network using memristive devices, so called neuromorphic system, has been extensively studied. Especially, on-chip (in situ) learning methods where training occurs inside hardware structure itself have been proposed and optimized based on memristor crossbar arrays regarding the linearity of weight-update characteristics. In this study, we analyze the effect of conductance overlap region of memristor on the recognition accuracy for on-chip learning simulation. The effect of conductance overlap region on recognition accuracy for modified national institute of standards and technology (MNIST) dataset is studied with an identical potentiation/depression pulse applied to Pt/Al2O3/TiOx/Ti/Pt stacked memristor. The overlap range can be varied by different pulse amplitude, and the training characteristics of memristive neural network is significantly dependent on the weight-update overlap region.",

keywords = "Memristor, Neural network, Neuromorphic system, On-chip learning, Weight overlap region",

author = "Lee, {Geun Ho} and Kim, {Tae Hyeon} and Song, {Min Suk} and Jinwoo Park and Sungjoon Kim and Kyungho Hong and Yoon Kim and Park, {Byung Gook} and Hyungjin Kim",

note = "Publisher Copyright: {\textcopyright} 2022",

year = "2022",

month = apr,

doi = "10.1016/j.chaos.2022.111999",

language = "English",

volume = "157",

journal = "Chaos, Solitons and Fractals",

issn = "0960-0779",

publisher = "Elsevier Ltd.",

}

TY - JOUR

T1 - Effect of weight overlap region on neuromorphic system with memristive synaptic devices

AU - Lee, Geun Ho

AU - Kim, Tae Hyeon

AU - Song, Min Suk

AU - Park, Jinwoo

AU - Kim, Sungjoon

AU - Hong, Kyungho

AU - Kim, Yoon

AU - Park, Byung Gook

AU - Kim, Hyungjin

PY - 2022/4

Y1 - 2022/4

N2 - Recently, hardware-based neural network using memristive devices, so called neuromorphic system, has been extensively studied. Especially, on-chip (in situ) learning methods where training occurs inside hardware structure itself have been proposed and optimized based on memristor crossbar arrays regarding the linearity of weight-update characteristics. In this study, we analyze the effect of conductance overlap region of memristor on the recognition accuracy for on-chip learning simulation. The effect of conductance overlap region on recognition accuracy for modified national institute of standards and technology (MNIST) dataset is studied with an identical potentiation/depression pulse applied to Pt/Al2O3/TiOx/Ti/Pt stacked memristor. The overlap range can be varied by different pulse amplitude, and the training characteristics of memristive neural network is significantly dependent on the weight-update overlap region.

AB - Recently, hardware-based neural network using memristive devices, so called neuromorphic system, has been extensively studied. Especially, on-chip (in situ) learning methods where training occurs inside hardware structure itself have been proposed and optimized based on memristor crossbar arrays regarding the linearity of weight-update characteristics. In this study, we analyze the effect of conductance overlap region of memristor on the recognition accuracy for on-chip learning simulation. The effect of conductance overlap region on recognition accuracy for modified national institute of standards and technology (MNIST) dataset is studied with an identical potentiation/depression pulse applied to Pt/Al2O3/TiOx/Ti/Pt stacked memristor. The overlap range can be varied by different pulse amplitude, and the training characteristics of memristive neural network is significantly dependent on the weight-update overlap region.

KW - Memristor

KW - Neural network

KW - Neuromorphic system

KW - On-chip learning

KW - Weight overlap region

UR - http://www.scopus.com/inward/record.url?scp=85126363825&partnerID=8YFLogxK

U2 - 10.1016/j.chaos.2022.111999

DO - 10.1016/j.chaos.2022.111999

M3 - Article

AN - SCOPUS:85126363825

SN - 0960-0779

VL - 157

JO - Chaos, Solitons and Fractals

JF - Chaos, Solitons and Fractals

M1 - 111999

ER -

Effect of weight overlap region on neuromorphic system with memristive synaptic devices

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this