Evaluating the Impact of Fault-Tolerance Capability of Deep Neural Networks Caused by Faults

Yung Yu Tsai; Jin Fu Li

doi:10.1109/SOCC52499.2021.9739383

Evaluating the Impact of Fault-Tolerance Capability of Deep Neural Networks Caused by Faults

Yung Yu Tsai, Jin Fu Li

電機工程學系

研究成果: 書貢獻/報告類型 › 會議論文篇章 › 同行評審

6 引文斯高帕斯（Scopus）

摘要

Deep neural network (DNN) is one effective technique used for the artificial intelligence applications. A DNN consists of a large amount of neurons arranged in a form of multilayers. Typically, a DNN has over-provisioning neurons such that it has inherent fault-tolerance capability. However, how to evaluate the fault-tolerance capability of DNNs is a question. In this paper, we propose a simulator to estimate the loss of inference accuracy due to the faults in a DNN model or the memory faults in hardware accelerator. The simulator is implemented based on the platforms of Keras and Tensorflow. It can evaluate the fault-tolerance capability of a DNN at model and hardware levels.

原文	???core.languages.en_GB???
主出版物標題	Proceedings - 34th IEEE International System-on-Chip Conference, SOCC 2021
編輯	Gang Qu, Jinjun Xiong, Danella Zhao, Venki Muthukumar, Md Farhadur Reza, Ramalingam Sridhar
發行者	IEEE Computer Society
頁面	272-277
頁數	6
ISBN（電子）	9781665429313
DOIs	https://doi.org/10.1109/SOCC52499.2021.9739383
出版狀態	已出版 - 2021
事件	34th IEEE International System-on-Chip Conference, SOCC 2021 - Virtual, Online, United States 持續時間: 14 9月 2021 → 17 9月 2021

出版系列

名字	International System on Chip Conference
卷	2021-September
ISSN（列印）	2164-1676
ISSN（電子）	2164-1706

???event.eventtypes.event.conference???

???event.eventtypes.event.conference???	34th IEEE International System-on-Chip Conference, SOCC 2021
國家/地區	United States
城市	Virtual, Online
期間	14/09/21 → 17/09/21

存取文件

10.1109/SOCC52499.2021.9739383

深度神經網路系統之強健性與可靠性增強技術(2/3)
Li, J.-F. (PI)
1/08/20 → 31/07/21
研究計畫: Research
應用於人體姿勢辨識與機器人之可重組深度神經網路引擎－總計畫暨子計畫一：應用於監督式學習之可重組深度神經網路技術(1/3)(2/3)
Li, J.-F. (PI)
1/08/20 → 31/07/21
研究計畫: Research

引用此

Tsai, Y. Y., & Li, J. F. (2021). Evaluating the Impact of Fault-Tolerance Capability of Deep Neural Networks Caused by Faults. 於 G. Qu, J. Xiong, D. Zhao, V. Muthukumar, M. F. Reza, & R. Sridhar (編輯), Proceedings - 34th IEEE International System-on-Chip Conference, SOCC 2021 (頁 272-277). (International System on Chip Conference; 卷 2021-September). IEEE Computer Society. https://doi.org/10.1109/SOCC52499.2021.9739383

Tsai, Yung Yu ; Li, Jin Fu. / Evaluating the Impact of Fault-Tolerance Capability of Deep Neural Networks Caused by Faults. Proceedings - 34th IEEE International System-on-Chip Conference, SOCC 2021. 編輯 / Gang Qu ; Jinjun Xiong ; Danella Zhao ; Venki Muthukumar ; Md Farhadur Reza ; Ramalingam Sridhar. IEEE Computer Society, 2021. 頁 272-277 (International System on Chip Conference).

@inproceedings{3e77d50ae0054d688b8c40a02d3751c6,

title = "Evaluating the Impact of Fault-Tolerance Capability of Deep Neural Networks Caused by Faults",

abstract = "Deep neural network (DNN) is one effective technique used for the artificial intelligence applications. A DNN consists of a large amount of neurons arranged in a form of multilayers. Typically, a DNN has over-provisioning neurons such that it has inherent fault-tolerance capability. However, how to evaluate the fault-tolerance capability of DNNs is a question. In this paper, we propose a simulator to estimate the loss of inference accuracy due to the faults in a DNN model or the memory faults in hardware accelerator. The simulator is implemented based on the platforms of Keras and Tensorflow. It can evaluate the fault-tolerance capability of a DNN at model and hardware levels. ",

keywords = "Deep neural network, convolutional neural newtork, fault tolerance, simulator",

author = "Tsai, \{Yung Yu\} and Li, \{Jin Fu\}",

note = "Publisher Copyright: {\textcopyright} 2021 IEEE.; 34th IEEE International System-on-Chip Conference, SOCC 2021 ; Conference date: 14-09-2021 Through 17-09-2021",

year = "2021",

doi = "10.1109/SOCC52499.2021.9739383",

language = "???core.languages.en\_GB???",

series = "International System on Chip Conference",

publisher = "IEEE Computer Society",

pages = "272--277",

editor = "Gang Qu and Jinjun Xiong and Danella Zhao and Venki Muthukumar and Reza, \{Md Farhadur\} and Ramalingam Sridhar",

booktitle = "Proceedings - 34th IEEE International System-on-Chip Conference, SOCC 2021",

}

Tsai, YY & Li, JF 2021, Evaluating the Impact of Fault-Tolerance Capability of Deep Neural Networks Caused by Faults. 於 G Qu, J Xiong, D Zhao, V Muthukumar, MF Reza & R Sridhar (編輯), Proceedings - 34th IEEE International System-on-Chip Conference, SOCC 2021. International System on Chip Conference, 卷 2021-September, IEEE Computer Society, 頁 272-277, 34th IEEE International System-on-Chip Conference, SOCC 2021, Virtual, Online, United States, 14/09/21. https://doi.org/10.1109/SOCC52499.2021.9739383

Evaluating the Impact of Fault-Tolerance Capability of Deep Neural Networks Caused by Faults. / Tsai, Yung Yu; Li, Jin Fu.
Proceedings - 34th IEEE International System-on-Chip Conference, SOCC 2021. 編輯 / Gang Qu; Jinjun Xiong; Danella Zhao; Venki Muthukumar; Md Farhadur Reza; Ramalingam Sridhar. IEEE Computer Society, 2021. p. 272-277 (International System on Chip Conference; 卷 2021-September).

研究成果: 書貢獻/報告類型 › 會議論文篇章 › 同行評審

TY - GEN

T1 - Evaluating the Impact of Fault-Tolerance Capability of Deep Neural Networks Caused by Faults

AU - Tsai, Yung Yu

AU - Li, Jin Fu

PY - 2021

Y1 - 2021

N2 - Deep neural network (DNN) is one effective technique used for the artificial intelligence applications. A DNN consists of a large amount of neurons arranged in a form of multilayers. Typically, a DNN has over-provisioning neurons such that it has inherent fault-tolerance capability. However, how to evaluate the fault-tolerance capability of DNNs is a question. In this paper, we propose a simulator to estimate the loss of inference accuracy due to the faults in a DNN model or the memory faults in hardware accelerator. The simulator is implemented based on the platforms of Keras and Tensorflow. It can evaluate the fault-tolerance capability of a DNN at model and hardware levels.

AB - Deep neural network (DNN) is one effective technique used for the artificial intelligence applications. A DNN consists of a large amount of neurons arranged in a form of multilayers. Typically, a DNN has over-provisioning neurons such that it has inherent fault-tolerance capability. However, how to evaluate the fault-tolerance capability of DNNs is a question. In this paper, we propose a simulator to estimate the loss of inference accuracy due to the faults in a DNN model or the memory faults in hardware accelerator. The simulator is implemented based on the platforms of Keras and Tensorflow. It can evaluate the fault-tolerance capability of a DNN at model and hardware levels.

KW - Deep neural network

KW - convolutional neural newtork

KW - fault tolerance

KW - simulator

UR - https://www.scopus.com/pages/publications/85127821235

U2 - 10.1109/SOCC52499.2021.9739383

DO - 10.1109/SOCC52499.2021.9739383

M3 - 會議論文篇章

AN - SCOPUS:85127821235

T3 - International System on Chip Conference

SP - 272

EP - 277

BT - Proceedings - 34th IEEE International System-on-Chip Conference, SOCC 2021

A2 - Qu, Gang

A2 - Xiong, Jinjun

A2 - Zhao, Danella

A2 - Muthukumar, Venki

A2 - Reza, Md Farhadur

A2 - Sridhar, Ramalingam

PB - IEEE Computer Society

T2 - 34th IEEE International System-on-Chip Conference, SOCC 2021

Y2 - 14 September 2021 through 17 September 2021

ER -

Tsai YY, Li JF. Evaluating the Impact of Fault-Tolerance Capability of Deep Neural Networks Caused by Faults. 於 Qu G, Xiong J, Zhao D, Muthukumar V, Reza MF, Sridhar R, 編輯, Proceedings - 34th IEEE International System-on-Chip Conference, SOCC 2021. IEEE Computer Society. 2021. p. 272-277. (International System on Chip Conference). doi: 10.1109/SOCC52499.2021.9739383

Evaluating the Impact of Fault-Tolerance Capability of Deep Neural Networks Caused by Faults

摘要

出版系列

???event.eventtypes.event.conference???

存取文件

指紋

專案

深度神經網路系統之強健性與可靠性增強技術(2/3)

應用於人體姿勢辨識與機器人之可重組深度神經網路引擎－總計畫暨子計畫一： 應用於監督式學習之可重組深度神經網路技術(1/3)(2/3)

引用此

應用於人體姿勢辨識與機器人之可重組深度神經網路引擎－總計畫暨子計畫一：應用於監督式學習之可重組深度神經網路技術(1/3)(2/3)