NV-BNN: An accurate deep convolutional neural network based on binary STT-MRAM for adaptive ai edge

Chih Cheng Chang; Ming Hung Wu; Jia Wei Lin; Chun Hsien Li; Vivek Parmar; Heng Yuan Lee; Jeng Hua Wei; Shyh Shyuan Sheu; Manan Suri; Tian-Sheuan Chang; Tuo-Hung Hou

doi:10.1145/3316781.3317872

NV-BNN: An accurate deep convolutional neural network based on binary STT-MRAM for adaptive ai edge

Chih Cheng Chang, Ming Hung Wu, Jia Wei Lin, Chun Hsien Li, Vivek Parmar, Heng Yuan Lee, Jeng Hua Wei, Shyh Shyuan Sheu, Manan Suri, Tian-Sheuan Chang, Tuo-Hung Hou

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

16 Scopus citations

Abstract

Binary STT-MRAM is a highly anticipated embedded nonvolatile memory technology in advanced logic nodes < 28 nm. How to enable its in-memory computing (IMC) capability is critical for enhancing AI Edge. Based on the soon-available STTMRAM, we report the first binary deep convolutional neural network (NV-BNN) capable of both local and remote learning. Exploiting intrinsic cumulative switching probability, accurate online training of CIFAR-10 color images (∼ 90%) is realized using a relaxed endurance spec (switching ≤ 20 times) and hybrid digital/IMC design. For offline training, the accuracy loss due to imprecise weight placement can be mitigated using a rapid noniterative training-with-noise and fine-tuning scheme.

Original language	English
Title of host publication	Proceedings of the 56th Annual Design Automation Conference 2019, DAC 2019
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9781450367257
DOIs	https://doi.org/10.1145/3316781.3317872
State	Published - 2 Jun 2019
Event	56th Annual Design Automation Conference, DAC 2019 - Las Vegas, United States Duration: 2 Jun 2019 → 6 Jun 2019

Publication series

Name	Proceedings - Design Automation Conference
ISSN (Print)	0738-100X

Conference

Conference	56th Annual Design Automation Conference, DAC 2019
Country/Territory	United States
City	Las Vegas
Period	2/06/19 → 6/06/19

Access to Document

10.1145/3316781.3317872

Cite this

Chang, C. C., Wu, M. H., Lin, J. W., Li, C. H., Parmar, V., Lee, H. Y., Wei, J. H., Sheu, S. S., Suri, M., Chang, T.-S., & Hou, T.-H. (2019). NV-BNN: An accurate deep convolutional neural network based on binary STT-MRAM for adaptive ai edge. In Proceedings of the 56th Annual Design Automation Conference 2019, DAC 2019 Article a30 (Proceedings - Design Automation Conference). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1145/3316781.3317872

@inproceedings{37cc255131c0441a8852e2f3c3aefb19,

title = "NV-BNN: An accurate deep convolutional neural network based on binary STT-MRAM for adaptive ai edge",

abstract = "Binary STT-MRAM is a highly anticipated embedded nonvolatile memory technology in advanced logic nodes < 28 nm. How to enable its in-memory computing (IMC) capability is critical for enhancing AI Edge. Based on the soon-available STTMRAM, we report the first binary deep convolutional neural network (NV-BNN) capable of both local and remote learning. Exploiting intrinsic cumulative switching probability, accurate online training of CIFAR-10 color images (∼ 90%) is realized using a relaxed endurance spec (switching ≤ 20 times) and hybrid digital/IMC design. For offline training, the accuracy loss due to imprecise weight placement can be mitigated using a rapid noniterative training-with-noise and fine-tuning scheme.",

author = "Chang, {Chih Cheng} and Wu, {Ming Hung} and Lin, {Jia Wei} and Li, {Chun Hsien} and Vivek Parmar and Lee, {Heng Yuan} and Wei, {Jeng Hua} and Sheu, {Shyh Shyuan} and Manan Suri and Tian-Sheuan Chang and Tuo-Hung Hou",

year = "2019",

month = jun,

day = "2",

doi = "10.1145/3316781.3317872",

language = "English",

series = "Proceedings - Design Automation Conference",

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

booktitle = "Proceedings of the 56th Annual Design Automation Conference 2019, DAC 2019",

address = "美國",

note = "56th Annual Design Automation Conference, DAC 2019 ; Conference date: 02-06-2019 Through 06-06-2019",

}

Chang, CC, Wu, MH, Lin, JW, Li, CH, Parmar, V, Lee, HY, Wei, JH, Sheu, SS, Suri, M, Chang, T-S & Hou, T-H 2019, NV-BNN: An accurate deep convolutional neural network based on binary STT-MRAM for adaptive ai edge. in Proceedings of the 56th Annual Design Automation Conference 2019, DAC 2019., a30, Proceedings - Design Automation Conference, Institute of Electrical and Electronics Engineers Inc., 56th Annual Design Automation Conference, DAC 2019, Las Vegas, United States, 2/06/19. https://doi.org/10.1145/3316781.3317872

NV-BNN: An accurate deep convolutional neural network based on binary STT-MRAM for adaptive ai edge. / Chang, Chih Cheng; Wu, Ming Hung; Lin, Jia Wei et al.
Proceedings of the 56th Annual Design Automation Conference 2019, DAC 2019. Institute of Electrical and Electronics Engineers Inc., 2019. a30 (Proceedings - Design Automation Conference).

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

TY - GEN

T1 - NV-BNN

T2 - 56th Annual Design Automation Conference, DAC 2019

AU - Chang, Chih Cheng

AU - Wu, Ming Hung

AU - Lin, Jia Wei

AU - Li, Chun Hsien

AU - Parmar, Vivek

AU - Lee, Heng Yuan

AU - Wei, Jeng Hua

AU - Sheu, Shyh Shyuan

AU - Suri, Manan

AU - Chang, Tian-Sheuan

AU - Hou, Tuo-Hung

PY - 2019/6/2

Y1 - 2019/6/2

N2 - Binary STT-MRAM is a highly anticipated embedded nonvolatile memory technology in advanced logic nodes < 28 nm. How to enable its in-memory computing (IMC) capability is critical for enhancing AI Edge. Based on the soon-available STTMRAM, we report the first binary deep convolutional neural network (NV-BNN) capable of both local and remote learning. Exploiting intrinsic cumulative switching probability, accurate online training of CIFAR-10 color images (∼ 90%) is realized using a relaxed endurance spec (switching ≤ 20 times) and hybrid digital/IMC design. For offline training, the accuracy loss due to imprecise weight placement can be mitigated using a rapid noniterative training-with-noise and fine-tuning scheme.

AB - Binary STT-MRAM is a highly anticipated embedded nonvolatile memory technology in advanced logic nodes < 28 nm. How to enable its in-memory computing (IMC) capability is critical for enhancing AI Edge. Based on the soon-available STTMRAM, we report the first binary deep convolutional neural network (NV-BNN) capable of both local and remote learning. Exploiting intrinsic cumulative switching probability, accurate online training of CIFAR-10 color images (∼ 90%) is realized using a relaxed endurance spec (switching ≤ 20 times) and hybrid digital/IMC design. For offline training, the accuracy loss due to imprecise weight placement can be mitigated using a rapid noniterative training-with-noise and fine-tuning scheme.

UR - http://www.scopus.com/inward/record.url?scp=85067820697&partnerID=8YFLogxK

U2 - 10.1145/3316781.3317872

DO - 10.1145/3316781.3317872

M3 - Conference contribution

AN - SCOPUS:85067820697

T3 - Proceedings - Design Automation Conference

BT - Proceedings of the 56th Annual Design Automation Conference 2019, DAC 2019

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 2 June 2019 through 6 June 2019

ER -

Chang CC, Wu MH, Lin JW, Li CH, Parmar V, Lee HY et al. NV-BNN: An accurate deep convolutional neural network based on binary STT-MRAM for adaptive ai edge. In Proceedings of the 56th Annual Design Automation Conference 2019, DAC 2019. Institute of Electrical and Electronics Engineers Inc. 2019. a30. (Proceedings - Design Automation Conference). doi: 10.1145/3316781.3317872

NV-BNN: An accurate deep convolutional neural network based on binary STT-MRAM for adaptive ai edge

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this