Fuse Devices for Pruning in Memristive Neural Network

Tae Hyeon Kim; Kyungho Hong; Sungjoon Kim; Jinwoo Park; Sangwook Youn; Jong Ho Lee; Byung Gook Park; Hyungjin Kim; Woo Young Choi

doi:10.1109/LED.2023.3237651

Fuse Devices for Pruning in Memristive Neural Network

Tae Hyeon Kim, Kyungho Hong, Sungjoon Kim, Jinwoo Park, Sangwook Youn, Jong Ho Lee, Byung Gook Park, Hyungjin Kim, Woo Young Choi

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

4 Scopus citations

Abstract

In this study, we developed a fuse device for pruning implementation in a hardware neural network. A line-shaped fuse device was fabricated with aluminum metal and characterized in a one fuse and one resistive random-access memory (1F1R) structure. A blow time of 0.4 μ and read endurance of > 107 were achieved, and the cut-off operation of the fuse was successfully verified in 1F1R. In addition, we developed a fuse design method for blow voltage and current via adjustment of the length, width, and thickness of the fuse. The adjustments were adopted to utilize the fuse in various synaptic devices. Finally, using simulations, we demonstrated a performance improvement due to the network pruning wherein the defective devices are disconnected by the fuse operations.

Original language	English
Pages (from-to)	520-523
Number of pages	4
Journal	IEEE Electron Device Letters
Volume	44
Issue number	3
DOIs	https://doi.org/10.1109/LED.2023.3237651
State	Published - 1 Mar 2023

Bibliographical note

Publisher Copyright:
© 1980-2012 IEEE.

Keywords

Fuse device
hardware neural network
network pruning
neuromorphic system
synaptic device

Access to Document

10.1109/LED.2023.3237651

Cite this

@article{497653f306cb49f18966c32347eec8f4,

title = "Fuse Devices for Pruning in Memristive Neural Network",

abstract = "In this study, we developed a fuse device for pruning implementation in a hardware neural network. A line-shaped fuse device was fabricated with aluminum metal and characterized in a one fuse and one resistive random-access memory (1F1R) structure. A blow time of 0.4 μ and read endurance of > 107 were achieved, and the cut-off operation of the fuse was successfully verified in 1F1R. In addition, we developed a fuse design method for blow voltage and current via adjustment of the length, width, and thickness of the fuse. The adjustments were adopted to utilize the fuse in various synaptic devices. Finally, using simulations, we demonstrated a performance improvement due to the network pruning wherein the defective devices are disconnected by the fuse operations.",

keywords = "Fuse device, hardware neural network, network pruning, neuromorphic system, synaptic device",

author = "Kim, {Tae Hyeon} and Kyungho Hong and Sungjoon Kim and Jinwoo Park and Sangwook Youn and Lee, {Jong Ho} and Park, {Byung Gook} and Hyungjin Kim and Choi, {Woo Young}",

note = "Publisher Copyright: {\textcopyright} 1980-2012 IEEE.",

year = "2023",

month = mar,

day = "1",

doi = "10.1109/LED.2023.3237651",

language = "English",

volume = "44",

pages = "520--523",

journal = "IEEE Electron Device Letters",

issn = "0741-3106",

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

number = "3",

}

TY - JOUR

T1 - Fuse Devices for Pruning in Memristive Neural Network

AU - Kim, Tae Hyeon

AU - Hong, Kyungho

AU - Kim, Sungjoon

AU - Park, Jinwoo

AU - Youn, Sangwook

AU - Lee, Jong Ho

AU - Park, Byung Gook

AU - Kim, Hyungjin

AU - Choi, Woo Young

PY - 2023/3/1

Y1 - 2023/3/1

N2 - In this study, we developed a fuse device for pruning implementation in a hardware neural network. A line-shaped fuse device was fabricated with aluminum metal and characterized in a one fuse and one resistive random-access memory (1F1R) structure. A blow time of 0.4 μ and read endurance of > 107 were achieved, and the cut-off operation of the fuse was successfully verified in 1F1R. In addition, we developed a fuse design method for blow voltage and current via adjustment of the length, width, and thickness of the fuse. The adjustments were adopted to utilize the fuse in various synaptic devices. Finally, using simulations, we demonstrated a performance improvement due to the network pruning wherein the defective devices are disconnected by the fuse operations.

AB - In this study, we developed a fuse device for pruning implementation in a hardware neural network. A line-shaped fuse device was fabricated with aluminum metal and characterized in a one fuse and one resistive random-access memory (1F1R) structure. A blow time of 0.4 μ and read endurance of > 107 were achieved, and the cut-off operation of the fuse was successfully verified in 1F1R. In addition, we developed a fuse design method for blow voltage and current via adjustment of the length, width, and thickness of the fuse. The adjustments were adopted to utilize the fuse in various synaptic devices. Finally, using simulations, we demonstrated a performance improvement due to the network pruning wherein the defective devices are disconnected by the fuse operations.

KW - Fuse device

KW - hardware neural network

KW - network pruning

KW - neuromorphic system

KW - synaptic device

UR - http://www.scopus.com/inward/record.url?scp=85147266701&partnerID=8YFLogxK

U2 - 10.1109/LED.2023.3237651

DO - 10.1109/LED.2023.3237651

M3 - Article

AN - SCOPUS:85147266701

SN - 0741-3106

VL - 44

SP - 520

EP - 523

JO - IEEE Electron Device Letters

JF - IEEE Electron Device Letters

IS - 3

ER -

Fuse Devices for Pruning in Memristive Neural Network

Abstract

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this