Cross-layer approaches for improving the dependability of deep learning systems

Muhammad Abdullah Hanif; Le Ha Hoang; Muhammad Shafique

doi:10.1145/3378678.3391884

Cross-layer approaches for improving the dependability of deep learning systems

Muhammad Abdullah Hanif, Le Ha Hoang, Muhammad Shafique

Electrical Engineering

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

Abstract

Deep Neural Networks (DNNs) - the state-of-the-art computational models for many Artificial Intelligence (AI) applications - are inherently compute and resource-intensive and, hence, cannot exploit traditional redundancy-based fault mitigation techniques for enhancing the dependability of DNN-based systems. Therefore, there is a dire need to search for alternate methods that can improve their reliability without high expenditure of resources by exploiting the intrinsic characteristics of these networks. In this paper, we present cross-layer approaches that, based on the intrinsic characteristics of DNNs, employ software and hardware-level modifications for improving the resilience of DNN-based systems to hardware-level faults, e.g., soft errors and permanent faults.

Original language	English (US)
Title of host publication	Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020
Editors	Sander Stuijk
Publisher	Association for Computing Machinery, Inc
Pages	78-81
Number of pages	4
ISBN (Electronic)	9781450371315
DOIs	https://doi.org/10.1145/3378678.3391884
State	Published - May 25 2020
Event	23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020 - St. Goar, Germany Duration: May 25 2020 → May 26 2020

Publication series

Name	Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020

Conference

Conference	23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020
Country	Germany
City	St. Goar
Period	5/25/20 → 5/26/20

Keywords

cross-layer
deep learning
dependability
DNNs
faults
hardware accelerators
neural networks
reliability
soft errors
systems

ASJC Scopus subject areas

Software
Hardware and Architecture

Access to Document

10.1145/3378678.3391884

Fingerprint Dive into the research topics of 'Cross-layer approaches for improving the dependability of deep learning systems'. Together they form a unique fingerprint.

Cite this

Hanif, M. A., Hoang, L. H., & Shafique, M. (2020). Cross-layer approaches for improving the dependability of deep learning systems. In S. Stuijk (Ed.), Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020 (pp. 78-81). (Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020). Association for Computing Machinery, Inc. https://doi.org/10.1145/3378678.3391884

Cross-layer approaches for improving the dependability of deep learning systems. / Hanif, Muhammad Abdullah; Hoang, Le Ha; Shafique, Muhammad.

Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020. ed. / Sander Stuijk. Association for Computing Machinery, Inc, 2020. p. 78-81 (Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020).

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

Hanif, MA, Hoang, LH & Shafique, M 2020, Cross-layer approaches for improving the dependability of deep learning systems. in S Stuijk (ed.), Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020. Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020, Association for Computing Machinery, Inc, pp. 78-81, 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020, St. Goar, Germany, 5/25/20. https://doi.org/10.1145/3378678.3391884

Hanif MA, Hoang LH, Shafique M. Cross-layer approaches for improving the dependability of deep learning systems. In Stuijk S, editor, Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020. Association for Computing Machinery, Inc. 2020. p. 78-81. (Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020). https://doi.org/10.1145/3378678.3391884

Hanif, Muhammad Abdullah ; Hoang, Le Ha ; Shafique, Muhammad. / Cross-layer approaches for improving the dependability of deep learning systems. Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020. editor / Sander Stuijk. Association for Computing Machinery, Inc, 2020. pp. 78-81 (Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020).

@inproceedings{843bb605bcbb44bc98cca474091fbebf,

title = "Cross-layer approaches for improving the dependability of deep learning systems",

abstract = "Deep Neural Networks (DNNs) - the state-of-the-art computational models for many Artificial Intelligence (AI) applications - are inherently compute and resource-intensive and, hence, cannot exploit traditional redundancy-based fault mitigation techniques for enhancing the dependability of DNN-based systems. Therefore, there is a dire need to search for alternate methods that can improve their reliability without high expenditure of resources by exploiting the intrinsic characteristics of these networks. In this paper, we present cross-layer approaches that, based on the intrinsic characteristics of DNNs, employ software and hardware-level modifications for improving the resilience of DNN-based systems to hardware-level faults, e.g., soft errors and permanent faults.",

keywords = "cross-layer, deep learning, dependability, DNNs, faults, hardware accelerators, neural networks, reliability, soft errors, systems",

author = "Hanif, {Muhammad Abdullah} and Hoang, {Le Ha} and Muhammad Shafique",

note = "Publisher Copyright: {\textcopyright} 2020 ACM. Copyright: Copyright 2020 Elsevier B.V., All rights reserved.; 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020 ; Conference date: 25-05-2020 Through 26-05-2020",

year = "2020",

month = may,

day = "25",

doi = "10.1145/3378678.3391884",

language = "English (US)",

series = "Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020",

publisher = "Association for Computing Machinery, Inc",

pages = "78--81",

editor = "Sander Stuijk",

booktitle = "Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020",

}

TY - GEN

T1 - Cross-layer approaches for improving the dependability of deep learning systems

AU - Hanif, Muhammad Abdullah

AU - Hoang, Le Ha

AU - Shafique, Muhammad

PY - 2020/5/25

Y1 - 2020/5/25

N2 - Deep Neural Networks (DNNs) - the state-of-the-art computational models for many Artificial Intelligence (AI) applications - are inherently compute and resource-intensive and, hence, cannot exploit traditional redundancy-based fault mitigation techniques for enhancing the dependability of DNN-based systems. Therefore, there is a dire need to search for alternate methods that can improve their reliability without high expenditure of resources by exploiting the intrinsic characteristics of these networks. In this paper, we present cross-layer approaches that, based on the intrinsic characteristics of DNNs, employ software and hardware-level modifications for improving the resilience of DNN-based systems to hardware-level faults, e.g., soft errors and permanent faults.

AB - Deep Neural Networks (DNNs) - the state-of-the-art computational models for many Artificial Intelligence (AI) applications - are inherently compute and resource-intensive and, hence, cannot exploit traditional redundancy-based fault mitigation techniques for enhancing the dependability of DNN-based systems. Therefore, there is a dire need to search for alternate methods that can improve their reliability without high expenditure of resources by exploiting the intrinsic characteristics of these networks. In this paper, we present cross-layer approaches that, based on the intrinsic characteristics of DNNs, employ software and hardware-level modifications for improving the resilience of DNN-based systems to hardware-level faults, e.g., soft errors and permanent faults.

KW - cross-layer

KW - deep learning

KW - dependability

KW - DNNs

KW - faults

KW - hardware accelerators

KW - neural networks

KW - reliability

KW - soft errors

KW - systems

UR - http://www.scopus.com/inward/record.url?scp=85086735385&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85086735385&partnerID=8YFLogxK

U2 - 10.1145/3378678.3391884

DO - 10.1145/3378678.3391884

M3 - Conference contribution

AN - SCOPUS:85086735385

T3 - Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020

SP - 78

EP - 81

BT - Proceedings of the 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020

A2 - Stuijk, Sander

PB - Association for Computing Machinery, Inc

T2 - 23rd International Workshop on Software and Compilers for Embedded Systems, SCOPES 2020

Y2 - 25 May 2020 through 26 May 2020

ER -