Reliability Improvement in RRAM-based DNN for Edge Computing

Md Oli-Uz-Zaman; Saleh Ahmad Khan; Geng Yuan; Yanzhi Wang; Zhiheng Liao; Jingyan Fu; Caiwen Ding; Jinhui Wang

doi:10.1109/ISCAS48785.2022.9937260

Reliability Improvement in RRAM-based DNN for Edge Computing

Md Oli-Uz-Zaman, Saleh Ahmad Khan, Geng Yuan, Yanzhi Wang, Zhiheng Liao, Jingyan Fu, Caiwen Ding, Jinhui Wang

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

5 Scopus citations

Abstract

Recently, the Resistive Random Access Memory (RRAM) has been paid more attention for edge computing applications in both academia and industry, because it offers power efficiency and low latency to perform the complex analog in-situ matrix-vector multiplication-the most fundamental operation of Deep Neural Networks (DNNs). But the Stuck at Fault (SAF) defect makes the RRAM unreliable for the practical implementation. A differential mapping method (DMM) is proposed in this paper to improve reliability by mitigate SAF defects from RRAM-based DNNs. Firstly, the weight distribution for the VGG8 model with the CIFAR10 dataset is presented and analyzed. Then the DMM is used for recovering the inference accuracies at 0.1% to 50% SAFs. The experiment results show that the DMM can recover DNNs to their original inference accuracies (90%), when the ratio of SAFs is smaller than 7.5%. And even when the SAF is in the extreme condition 50%, it is still highly efficient to recover the inference accuracy to 80%. What is more, the DMM is a highly reliable regulator to avoid power and timing overhead generated by SAFs.

Original language	English (US)
Title of host publication	IEEE International Symposium on Circuits and Systems, ISCAS 2022
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	581-585
Number of pages	5
ISBN (Electronic)	9781665484855
DOIs	https://doi.org/10.1109/ISCAS48785.2022.9937260
State	Published - 2022
Externally published	Yes
Event	2022 IEEE International Symposium on Circuits and Systems, ISCAS 2022 - Austin, United States Duration: May 27 2022 → Jun 1 2022

Publication series

Name	Proceedings - IEEE International Symposium on Circuits and Systems
Volume	2022-May
ISSN (Print)	0271-4310

Conference

Conference	2022 IEEE International Symposium on Circuits and Systems, ISCAS 2022
Country/Territory	United States
City	Austin
Period	5/27/22 → 6/1/22

Bibliographical note

Publisher Copyright:
© 2022 IEEE.

Keywords

deep neural network (DNN)
differential mapping method
edge computing
latency
power
resistive random access memory (RRAM)
stuck at fault (SAF)

Publisher link

10.1109/ISCAS48785.2022.9937260

Find It

Find It at U of M Libraries

Cite this

Oli-Uz-Zaman, M., Khan, S. A., Yuan, G., Wang, Y., Liao, Z., Fu, J., Ding, C., & Wang, J. (2022). Reliability Improvement in RRAM-based DNN for Edge Computing. In IEEE International Symposium on Circuits and Systems, ISCAS 2022 (pp. 581-585). (Proceedings - IEEE International Symposium on Circuits and Systems; Vol. 2022-May). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ISCAS48785.2022.9937260

Reliability Improvement in RRAM-based DNN for Edge Computing. / Oli-Uz-Zaman, Md; Khan, Saleh Ahmad; Yuan, Geng et al.
IEEE International Symposium on Circuits and Systems, ISCAS 2022. Institute of Electrical and Electronics Engineers Inc., 2022. p. 581-585 (Proceedings - IEEE International Symposium on Circuits and Systems; Vol. 2022-May).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Oli-Uz-Zaman, M, Khan, SA, Yuan, G, Wang, Y, Liao, Z, Fu, J, Ding, C & Wang, J 2022, Reliability Improvement in RRAM-based DNN for Edge Computing. in IEEE International Symposium on Circuits and Systems, ISCAS 2022. Proceedings - IEEE International Symposium on Circuits and Systems, vol. 2022-May, Institute of Electrical and Electronics Engineers Inc., pp. 581-585, 2022 IEEE International Symposium on Circuits and Systems, ISCAS 2022, Austin, United States, 5/27/22. https://doi.org/10.1109/ISCAS48785.2022.9937260

Oli-Uz-Zaman M, Khan SA, Yuan G, Wang Y, Liao Z, Fu J et al. Reliability Improvement in RRAM-based DNN for Edge Computing. In IEEE International Symposium on Circuits and Systems, ISCAS 2022. Institute of Electrical and Electronics Engineers Inc. 2022. p. 581-585. (Proceedings - IEEE International Symposium on Circuits and Systems). doi: 10.1109/ISCAS48785.2022.9937260

@inproceedings{fe237e20084e4bbf98fc8a3cd8e477a7,

title = "Reliability Improvement in RRAM-based DNN for Edge Computing",

abstract = "Recently, the Resistive Random Access Memory (RRAM) has been paid more attention for edge computing applications in both academia and industry, because it offers power efficiency and low latency to perform the complex analog in-situ matrix-vector multiplication-the most fundamental operation of Deep Neural Networks (DNNs). But the Stuck at Fault (SAF) defect makes the RRAM unreliable for the practical implementation. A differential mapping method (DMM) is proposed in this paper to improve reliability by mitigate SAF defects from RRAM-based DNNs. Firstly, the weight distribution for the VGG8 model with the CIFAR10 dataset is presented and analyzed. Then the DMM is used for recovering the inference accuracies at 0.1% to 50% SAFs. The experiment results show that the DMM can recover DNNs to their original inference accuracies (90%), when the ratio of SAFs is smaller than 7.5%. And even when the SAF is in the extreme condition 50%, it is still highly efficient to recover the inference accuracy to 80%. What is more, the DMM is a highly reliable regulator to avoid power and timing overhead generated by SAFs.",

keywords = "deep neural network (DNN), differential mapping method, edge computing, latency, power, resistive random access memory (RRAM), stuck at fault (SAF)",

author = "Md Oli-Uz-Zaman and Khan, {Saleh Ahmad} and Geng Yuan and Yanzhi Wang and Zhiheng Liao and Jingyan Fu and Caiwen Ding and Jinhui Wang",

note = "Publisher Copyright: {\textcopyright} 2022 IEEE.; 2022 IEEE International Symposium on Circuits and Systems, ISCAS 2022 ; Conference date: 27-05-2022 Through 01-06-2022",

year = "2022",

doi = "10.1109/ISCAS48785.2022.9937260",

language = "English (US)",

series = "Proceedings - IEEE International Symposium on Circuits and Systems",

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

pages = "581--585",

booktitle = "IEEE International Symposium on Circuits and Systems, ISCAS 2022",

}

TY - GEN

T1 - Reliability Improvement in RRAM-based DNN for Edge Computing

AU - Oli-Uz-Zaman, Md

AU - Khan, Saleh Ahmad

AU - Yuan, Geng

AU - Wang, Yanzhi

AU - Liao, Zhiheng

AU - Fu, Jingyan

AU - Ding, Caiwen

AU - Wang, Jinhui

PY - 2022

Y1 - 2022

N2 - Recently, the Resistive Random Access Memory (RRAM) has been paid more attention for edge computing applications in both academia and industry, because it offers power efficiency and low latency to perform the complex analog in-situ matrix-vector multiplication-the most fundamental operation of Deep Neural Networks (DNNs). But the Stuck at Fault (SAF) defect makes the RRAM unreliable for the practical implementation. A differential mapping method (DMM) is proposed in this paper to improve reliability by mitigate SAF defects from RRAM-based DNNs. Firstly, the weight distribution for the VGG8 model with the CIFAR10 dataset is presented and analyzed. Then the DMM is used for recovering the inference accuracies at 0.1% to 50% SAFs. The experiment results show that the DMM can recover DNNs to their original inference accuracies (90%), when the ratio of SAFs is smaller than 7.5%. And even when the SAF is in the extreme condition 50%, it is still highly efficient to recover the inference accuracy to 80%. What is more, the DMM is a highly reliable regulator to avoid power and timing overhead generated by SAFs.

AB - Recently, the Resistive Random Access Memory (RRAM) has been paid more attention for edge computing applications in both academia and industry, because it offers power efficiency and low latency to perform the complex analog in-situ matrix-vector multiplication-the most fundamental operation of Deep Neural Networks (DNNs). But the Stuck at Fault (SAF) defect makes the RRAM unreliable for the practical implementation. A differential mapping method (DMM) is proposed in this paper to improve reliability by mitigate SAF defects from RRAM-based DNNs. Firstly, the weight distribution for the VGG8 model with the CIFAR10 dataset is presented and analyzed. Then the DMM is used for recovering the inference accuracies at 0.1% to 50% SAFs. The experiment results show that the DMM can recover DNNs to their original inference accuracies (90%), when the ratio of SAFs is smaller than 7.5%. And even when the SAF is in the extreme condition 50%, it is still highly efficient to recover the inference accuracy to 80%. What is more, the DMM is a highly reliable regulator to avoid power and timing overhead generated by SAFs.

KW - deep neural network (DNN)

KW - differential mapping method

KW - edge computing

KW - latency

KW - power

KW - resistive random access memory (RRAM)

KW - stuck at fault (SAF)

UR - http://www.scopus.com/inward/record.url?scp=85142507070&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85142507070&partnerID=8YFLogxK

U2 - 10.1109/ISCAS48785.2022.9937260

DO - 10.1109/ISCAS48785.2022.9937260

M3 - Conference contribution

AN - SCOPUS:85142507070

T3 - Proceedings - IEEE International Symposium on Circuits and Systems

SP - 581

EP - 585

BT - IEEE International Symposium on Circuits and Systems, ISCAS 2022

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2022 IEEE International Symposium on Circuits and Systems, ISCAS 2022

Y2 - 27 May 2022 through 1 June 2022

ER -

Reliability Improvement in RRAM-based DNN for Edge Computing

Abstract

Publication series

Conference

Bibliographical note

Keywords

Publisher link

Find It

Other files and links

Fingerprint

Cite this