RAM: Exploiting Restrained and Approximate Management for Enabling Neural Network Training on NVM-based Systems

Chien Chung Ho; Wei Chen Wang; Szu Yu Chen; Yung Chun Li; Kun Chi Chiang

doi:10.1145/3477314.3507090

RAM: Exploiting Restrained and Approximate Management for Enabling Neural Network Training on NVM-based Systems

Chien Chung Ho, Wei Chen Wang, Szu Yu Chen, Yung Chun Li, Kun Chi Chiang

MS Degree in AI Robotics

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

2 Citations (Scopus)

Abstract

Training neural networks (NN) over conventional DRAM-based edge devices are highly restricted due to DRAM's leakage-power and limited-density properties. The non-volatile memory (NVM) reveals its potential of providing solutions with the high-density and nearly-zero leakage power features; however, it could lead to severe performance and lifetime issues. This paper focuses on exploring the NVM-aware neural network training design to mitigate performance and lifetime issues caused by the conventional NN approaches. Especially, we aim at exploiting the restrained-and-approximate managements to leverage insignificant data on performance and lifetime optimization. The encouraging results were observed with the comparable validation accuracy in the evaluations.

Original language	English
Title of host publication	Proceedings of the 37th ACM/SIGAPP Symposium on Applied Computing, SAC 2022
Publisher	Association for Computing Machinery
Pages	116-123
Number of pages	8
ISBN (Electronic)	9781450387132
DOIs	https://doi.org/10.1145/3477314.3507090
Publication status	Published - 2022 Apr 25
Event	37th ACM/SIGAPP Symposium on Applied Computing, SAC 2022 - Virtual, Online Duration: 2022 Apr 25 → 2022 Apr 29

Publication series

Name	Proceedings of the ACM Symposium on Applied Computing

Conference

Conference	37th ACM/SIGAPP Symposium on Applied Computing, SAC 2022
City	Virtual, Online
Period	22-04-25 → 22-04-29

All Science Journal Classification (ASJC) codes

Software

Access to Document

10.1145/3477314.3507090

Cite this

Ho, C. C., Wang, W. C., Chen, S. Y., Li, Y. C., & Chiang, K. C. (2022). RAM: Exploiting Restrained and Approximate Management for Enabling Neural Network Training on NVM-based Systems. In Proceedings of the 37th ACM/SIGAPP Symposium on Applied Computing, SAC 2022 (pp. 116-123). (Proceedings of the ACM Symposium on Applied Computing). Association for Computing Machinery. https://doi.org/10.1145/3477314.3507090

@inproceedings{de79dafdcc9d4ac5bcf9a151e930dbeb,

title = "RAM: Exploiting Restrained and Approximate Management for Enabling Neural Network Training on NVM-based Systems",

abstract = "Training neural networks (NN) over conventional DRAM-based edge devices are highly restricted due to DRAM's leakage-power and limited-density properties. The non-volatile memory (NVM) reveals its potential of providing solutions with the high-density and nearly-zero leakage power features; however, it could lead to severe performance and lifetime issues. This paper focuses on exploring the NVM-aware neural network training design to mitigate performance and lifetime issues caused by the conventional NN approaches. Especially, we aim at exploiting the restrained-and-approximate managements to leverage insignificant data on performance and lifetime optimization. The encouraging results were observed with the comparable validation accuracy in the evaluations.",

author = "Ho, {Chien Chung} and Wang, {Wei Chen} and Chen, {Szu Yu} and Li, {Yung Chun} and Chiang, {Kun Chi}",

note = "Funding Information: This work was supported in part by the Ministry of Science and Technology, Taiwan under grant no. 109-2221-E-006-215-MY3. Publisher Copyright: {\textcopyright} 2022 ACM.; 37th ACM/SIGAPP Symposium on Applied Computing, SAC 2022 ; Conference date: 25-04-2022 Through 29-04-2022",

year = "2022",

month = apr,

day = "25",

doi = "10.1145/3477314.3507090",

language = "English",

series = "Proceedings of the ACM Symposium on Applied Computing",

publisher = "Association for Computing Machinery",

pages = "116--123",

booktitle = "Proceedings of the 37th ACM/SIGAPP Symposium on Applied Computing, SAC 2022",

}

Ho, CC, Wang, WC, Chen, SY, Li, YC & Chiang, KC 2022, RAM: Exploiting Restrained and Approximate Management for Enabling Neural Network Training on NVM-based Systems. in Proceedings of the 37th ACM/SIGAPP Symposium on Applied Computing, SAC 2022. Proceedings of the ACM Symposium on Applied Computing, Association for Computing Machinery, pp. 116-123, 37th ACM/SIGAPP Symposium on Applied Computing, SAC 2022, Virtual, Online, 22-04-25. https://doi.org/10.1145/3477314.3507090

RAM: Exploiting Restrained and Approximate Management for Enabling Neural Network Training on NVM-based Systems. / Ho, Chien Chung; Wang, Wei Chen; Chen, Szu Yu et al.
Proceedings of the 37th ACM/SIGAPP Symposium on Applied Computing, SAC 2022. Association for Computing Machinery, 2022. p. 116-123 (Proceedings of the ACM Symposium on Applied Computing).

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

TY - GEN

T1 - RAM

T2 - 37th ACM/SIGAPP Symposium on Applied Computing, SAC 2022

AU - Ho, Chien Chung

AU - Wang, Wei Chen

AU - Chen, Szu Yu

AU - Li, Yung Chun

AU - Chiang, Kun Chi

PY - 2022/4/25

Y1 - 2022/4/25

N2 - Training neural networks (NN) over conventional DRAM-based edge devices are highly restricted due to DRAM's leakage-power and limited-density properties. The non-volatile memory (NVM) reveals its potential of providing solutions with the high-density and nearly-zero leakage power features; however, it could lead to severe performance and lifetime issues. This paper focuses on exploring the NVM-aware neural network training design to mitigate performance and lifetime issues caused by the conventional NN approaches. Especially, we aim at exploiting the restrained-and-approximate managements to leverage insignificant data on performance and lifetime optimization. The encouraging results were observed with the comparable validation accuracy in the evaluations.

AB - Training neural networks (NN) over conventional DRAM-based edge devices are highly restricted due to DRAM's leakage-power and limited-density properties. The non-volatile memory (NVM) reveals its potential of providing solutions with the high-density and nearly-zero leakage power features; however, it could lead to severe performance and lifetime issues. This paper focuses on exploring the NVM-aware neural network training design to mitigate performance and lifetime issues caused by the conventional NN approaches. Especially, we aim at exploiting the restrained-and-approximate managements to leverage insignificant data on performance and lifetime optimization. The encouraging results were observed with the comparable validation accuracy in the evaluations.

UR - http://www.scopus.com/inward/record.url?scp=85130352921&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85130352921&partnerID=8YFLogxK

U2 - 10.1145/3477314.3507090

DO - 10.1145/3477314.3507090

M3 - Conference contribution

AN - SCOPUS:85130352921

T3 - Proceedings of the ACM Symposium on Applied Computing

SP - 116

EP - 123

BT - Proceedings of the 37th ACM/SIGAPP Symposium on Applied Computing, SAC 2022

PB - Association for Computing Machinery

Y2 - 25 April 2022 through 29 April 2022

ER -

Ho CC, Wang WC, Chen SY, Li YC, Chiang KC. RAM: Exploiting Restrained and Approximate Management for Enabling Neural Network Training on NVM-based Systems. In Proceedings of the 37th ACM/SIGAPP Symposium on Applied Computing, SAC 2022. Association for Computing Machinery. 2022. p. 116-123. (Proceedings of the ACM Symposium on Applied Computing). doi: 10.1145/3477314.3507090

RAM: Exploiting Restrained and Approximate Management for Enabling Neural Network Training on NVM-based Systems

Abstract

Publication series

Conference

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this