A 47.4µJ/epoch trainable deep convolutional neural network accelerator for in-situ personalization on smart devices

Seungkyu Choi; Jaehyeong Sim; Myeonggu Kang; Yeongjae Choi; Hyeonuk Kim; Lee Sup Kim

doi:10.1109/A-SSCC47793.2019.9056972

A 47.4µJ/epoch trainable deep convolutional neural network accelerator for in-situ personalization on smart devices

Seungkyu Choi, Jaehyeong Sim, Myeonggu Kang, Yeongjae Choi, Hyeonuk Kim, Lee Sup Kim

Computer Science & Engineering

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

2 Scopus citations

Abstract

A scalable deep learning accelerator supporting both inference and training is implemented for device personalization of deep convolutional neural networks. It consists of three processor cores operating with distinct energy-efficient dataflow for different types of computation in CNN training. Two cores conduct forward and backward propagation in convolutional layers and utilize a masking scheme to reduce 88.3% of intermediate data to store for training. The third core executes weight update process in convolutional layers and inner product computation in fully connected layers with a novel large window dataflow. The system enables 8-bit fixed point datapath with lossless training and consumes 47.4µJ/epoch for a customized deep CNN model.

Original language	English
Title of host publication	Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	57-60
Number of pages	4
ISBN (Electronic)	9781728151069
DOIs	https://doi.org/10.1109/A-SSCC47793.2019.9056972
State	Published - Nov 2019
Event	15th IEEE Asian Solid-State Circuits Conference, A-SSCC 2019 - Macao, China Duration: 4 Nov 2019 → 6 Nov 2019

Publication series

Name	Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019
Volume	2019-November

Conference

Conference	15th IEEE Asian Solid-State Circuits Conference, A-SSCC 2019
Country/Territory	China
City	Macao
Period	4/11/19 → 6/11/19

Access to Document

10.1109/A-SSCC47793.2019.9056972

Cite this

Choi, S., Sim, J., Kang, M., Choi, Y., Kim, H., & Kim, L. S. (2019). A 47.4µJ/epoch trainable deep convolutional neural network accelerator for in-situ personalization on smart devices. In Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019 (pp. 57-60). [9056972] (Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019; Vol. 2019-November). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/A-SSCC47793.2019.9056972

Choi, Seungkyu ; Sim, Jaehyeong ; Kang, Myeonggu et al. / A 47.4µJ/epoch trainable deep convolutional neural network accelerator for in-situ personalization on smart devices. Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019. Institute of Electrical and Electronics Engineers Inc., 2019. pp. 57-60 (Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019).

@inproceedings{24fd2214ab394da08d1d58f7553d25bf,

title = "A 47.4µJ/epoch trainable deep convolutional neural network accelerator for in-situ personalization on smart devices",

abstract = "A scalable deep learning accelerator supporting both inference and training is implemented for device personalization of deep convolutional neural networks. It consists of three processor cores operating with distinct energy-efficient dataflow for different types of computation in CNN training. Two cores conduct forward and backward propagation in convolutional layers and utilize a masking scheme to reduce 88.3% of intermediate data to store for training. The third core executes weight update process in convolutional layers and inner product computation in fully connected layers with a novel large window dataflow. The system enables 8-bit fixed point datapath with lossless training and consumes 47.4µJ/epoch for a customized deep CNN model.",

author = "Seungkyu Choi and Jaehyeong Sim and Myeonggu Kang and Yeongjae Choi and Hyeonuk Kim and Kim, {Lee Sup}",

note = "Publisher Copyright: {\textcopyright} 2019 Institute of Electrical and Electronics Engineers Inc.. All rights reserved.; 15th IEEE Asian Solid-State Circuits Conference, A-SSCC 2019 ; Conference date: 04-11-2019 Through 06-11-2019",

year = "2019",

month = nov,

doi = "10.1109/A-SSCC47793.2019.9056972",

language = "English",

series = "Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019",

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

pages = "57--60",

booktitle = "Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019",

}

Choi, S, Sim, J, Kang, M, Choi, Y, Kim, H & Kim, LS 2019, A 47.4µJ/epoch trainable deep convolutional neural network accelerator for in-situ personalization on smart devices. in Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019., 9056972, Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019, vol. 2019-November, Institute of Electrical and Electronics Engineers Inc., pp. 57-60, 15th IEEE Asian Solid-State Circuits Conference, A-SSCC 2019, Macao, China, 4/11/19. https://doi.org/10.1109/A-SSCC47793.2019.9056972

A 47.4µJ/epoch trainable deep convolutional neural network accelerator for in-situ personalization on smart devices. / Choi, Seungkyu; Sim, Jaehyeong; Kang, Myeonggu et al.
Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019. Institute of Electrical and Electronics Engineers Inc., 2019. p. 57-60 9056972 (Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019; Vol. 2019-November).

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

TY - GEN

T1 - A 47.4µJ/epoch trainable deep convolutional neural network accelerator for in-situ personalization on smart devices

AU - Choi, Seungkyu

AU - Sim, Jaehyeong

AU - Kang, Myeonggu

AU - Choi, Yeongjae

AU - Kim, Hyeonuk

AU - Kim, Lee Sup

PY - 2019/11

Y1 - 2019/11

N2 - A scalable deep learning accelerator supporting both inference and training is implemented for device personalization of deep convolutional neural networks. It consists of three processor cores operating with distinct energy-efficient dataflow for different types of computation in CNN training. Two cores conduct forward and backward propagation in convolutional layers and utilize a masking scheme to reduce 88.3% of intermediate data to store for training. The third core executes weight update process in convolutional layers and inner product computation in fully connected layers with a novel large window dataflow. The system enables 8-bit fixed point datapath with lossless training and consumes 47.4µJ/epoch for a customized deep CNN model.

AB - A scalable deep learning accelerator supporting both inference and training is implemented for device personalization of deep convolutional neural networks. It consists of three processor cores operating with distinct energy-efficient dataflow for different types of computation in CNN training. Two cores conduct forward and backward propagation in convolutional layers and utilize a masking scheme to reduce 88.3% of intermediate data to store for training. The third core executes weight update process in convolutional layers and inner product computation in fully connected layers with a novel large window dataflow. The system enables 8-bit fixed point datapath with lossless training and consumes 47.4µJ/epoch for a customized deep CNN model.

UR - http://www.scopus.com/inward/record.url?scp=85108985295&partnerID=8YFLogxK

U2 - 10.1109/A-SSCC47793.2019.9056972

DO - 10.1109/A-SSCC47793.2019.9056972

M3 - Conference contribution

AN - SCOPUS:85108985295

T3 - Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019

SP - 57

EP - 60

BT - Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 15th IEEE Asian Solid-State Circuits Conference, A-SSCC 2019

Y2 - 4 November 2019 through 6 November 2019

ER -

Choi S, Sim J, Kang M, Choi Y, Kim H, Kim LS. A 47.4µJ/epoch trainable deep convolutional neural network accelerator for in-situ personalization on smart devices. In Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019. Institute of Electrical and Electronics Engineers Inc. 2019. p. 57-60. 9056972. (Proceedings - 2019 IEEE Asian Solid-State Circuits Conference, A-SSCC 2019). doi: 10.1109/A-SSCC47793.2019.9056972

A 47.4µJ/epoch trainable deep convolutional neural network accelerator for in-situ personalization on smart devices

Abstract

Publication series

Conference

Access to Document

Fingerprint

Cite this