Exploring the capabilities of mobile devices in supporting deep learning

Yitao Chen, Saman Biookaghazadeh, Ming Zhao

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Deep neural networks (DNNs) have unleashed a new wave of applications on mobile devices, such as various intelligent personal assistants. Most of these applications rely on the use of cloud resources to perform deep learning. With increasingly more powerful mobile devices, users can perform more deep learning tasks on the devices. In addition, learning on the devices has important advantages, such as personalization, privacy, and responsiveness; however, a good understanding of the capabilities of modern mobile devices in supporting deep learning is generally lacking. To address this gap in knowledge, this paper presents a comprehensive study on performing training and inference on mobile devices. It develops TensorFlow+, an extension of the widely used TensorFlow framework, to enable training DNNs on devices and use the available GPUs to accelerate the learning tasks. The study focuses on four aspects: 1) the performance impact of the network architecture; 2) the effectiveness of using accelerators for learning on mobile devices; 3) the resource and battery usages of training and inference; and 4) the performance impact on other applications running on the devices. The results show that the size (width and depth) of a network as well as the types of layers that it uses are important to not only meeting the device's capability but also to the performance of learning. The study also shows that hardware acceleration is important to both improving the speed of learning and reducing the impact on other applications on the device.

Original languageEnglish (US)
Title of host publicationProceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019
PublisherAssociation for Computing Machinery, Inc
Pages127-138
Number of pages12
ISBN (Electronic)9781450367332
DOIs
StatePublished - Nov 7 2019
Event4th ACM/IEEE Symposium on Edge Computing, SEC 2019 - Arlington, United States
Duration: Nov 7 2019Nov 9 2019

Publication series

NameProceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019

Conference

Conference4th ACM/IEEE Symposium on Edge Computing, SEC 2019
CountryUnited States
CityArlington
Period11/7/1911/9/19

Fingerprint

Mobile devices
Intelligent agents
Network architecture
Particle accelerators
Deep learning
Hardware
Deep neural networks

Keywords

  • Deep learning
  • Edge computing
  • Mobile computing
  • Neural networks

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Hardware and Architecture

Cite this

Chen, Y., Biookaghazadeh, S., & Zhao, M. (2019). Exploring the capabilities of mobile devices in supporting deep learning. In Proceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019 (pp. 127-138). (Proceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019). Association for Computing Machinery, Inc. https://doi.org/10.1145/3318216.3363316

Exploring the capabilities of mobile devices in supporting deep learning. / Chen, Yitao; Biookaghazadeh, Saman; Zhao, Ming.

Proceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019. Association for Computing Machinery, Inc, 2019. p. 127-138 (Proceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Chen, Y, Biookaghazadeh, S & Zhao, M 2019, Exploring the capabilities of mobile devices in supporting deep learning. in Proceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019. Proceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019, Association for Computing Machinery, Inc, pp. 127-138, 4th ACM/IEEE Symposium on Edge Computing, SEC 2019, Arlington, United States, 11/7/19. https://doi.org/10.1145/3318216.3363316
Chen Y, Biookaghazadeh S, Zhao M. Exploring the capabilities of mobile devices in supporting deep learning. In Proceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019. Association for Computing Machinery, Inc. 2019. p. 127-138. (Proceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019). https://doi.org/10.1145/3318216.3363316
Chen, Yitao ; Biookaghazadeh, Saman ; Zhao, Ming. / Exploring the capabilities of mobile devices in supporting deep learning. Proceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019. Association for Computing Machinery, Inc, 2019. pp. 127-138 (Proceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019).
@inproceedings{9ab9dc25c3c44f89a5fad76245797920,
title = "Exploring the capabilities of mobile devices in supporting deep learning",
abstract = "Deep neural networks (DNNs) have unleashed a new wave of applications on mobile devices, such as various intelligent personal assistants. Most of these applications rely on the use of cloud resources to perform deep learning. With increasingly more powerful mobile devices, users can perform more deep learning tasks on the devices. In addition, learning on the devices has important advantages, such as personalization, privacy, and responsiveness; however, a good understanding of the capabilities of modern mobile devices in supporting deep learning is generally lacking. To address this gap in knowledge, this paper presents a comprehensive study on performing training and inference on mobile devices. It develops TensorFlow+, an extension of the widely used TensorFlow framework, to enable training DNNs on devices and use the available GPUs to accelerate the learning tasks. The study focuses on four aspects: 1) the performance impact of the network architecture; 2) the effectiveness of using accelerators for learning on mobile devices; 3) the resource and battery usages of training and inference; and 4) the performance impact on other applications running on the devices. The results show that the size (width and depth) of a network as well as the types of layers that it uses are important to not only meeting the device's capability but also to the performance of learning. The study also shows that hardware acceleration is important to both improving the speed of learning and reducing the impact on other applications on the device.",
keywords = "Deep learning, Edge computing, Mobile computing, Neural networks",
author = "Yitao Chen and Saman Biookaghazadeh and Ming Zhao",
year = "2019",
month = "11",
day = "7",
doi = "10.1145/3318216.3363316",
language = "English (US)",
series = "Proceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019",
publisher = "Association for Computing Machinery, Inc",
pages = "127--138",
booktitle = "Proceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019",

}

TY - GEN

T1 - Exploring the capabilities of mobile devices in supporting deep learning

AU - Chen, Yitao

AU - Biookaghazadeh, Saman

AU - Zhao, Ming

PY - 2019/11/7

Y1 - 2019/11/7

N2 - Deep neural networks (DNNs) have unleashed a new wave of applications on mobile devices, such as various intelligent personal assistants. Most of these applications rely on the use of cloud resources to perform deep learning. With increasingly more powerful mobile devices, users can perform more deep learning tasks on the devices. In addition, learning on the devices has important advantages, such as personalization, privacy, and responsiveness; however, a good understanding of the capabilities of modern mobile devices in supporting deep learning is generally lacking. To address this gap in knowledge, this paper presents a comprehensive study on performing training and inference on mobile devices. It develops TensorFlow+, an extension of the widely used TensorFlow framework, to enable training DNNs on devices and use the available GPUs to accelerate the learning tasks. The study focuses on four aspects: 1) the performance impact of the network architecture; 2) the effectiveness of using accelerators for learning on mobile devices; 3) the resource and battery usages of training and inference; and 4) the performance impact on other applications running on the devices. The results show that the size (width and depth) of a network as well as the types of layers that it uses are important to not only meeting the device's capability but also to the performance of learning. The study also shows that hardware acceleration is important to both improving the speed of learning and reducing the impact on other applications on the device.

AB - Deep neural networks (DNNs) have unleashed a new wave of applications on mobile devices, such as various intelligent personal assistants. Most of these applications rely on the use of cloud resources to perform deep learning. With increasingly more powerful mobile devices, users can perform more deep learning tasks on the devices. In addition, learning on the devices has important advantages, such as personalization, privacy, and responsiveness; however, a good understanding of the capabilities of modern mobile devices in supporting deep learning is generally lacking. To address this gap in knowledge, this paper presents a comprehensive study on performing training and inference on mobile devices. It develops TensorFlow+, an extension of the widely used TensorFlow framework, to enable training DNNs on devices and use the available GPUs to accelerate the learning tasks. The study focuses on four aspects: 1) the performance impact of the network architecture; 2) the effectiveness of using accelerators for learning on mobile devices; 3) the resource and battery usages of training and inference; and 4) the performance impact on other applications running on the devices. The results show that the size (width and depth) of a network as well as the types of layers that it uses are important to not only meeting the device's capability but also to the performance of learning. The study also shows that hardware acceleration is important to both improving the speed of learning and reducing the impact on other applications on the device.

KW - Deep learning

KW - Edge computing

KW - Mobile computing

KW - Neural networks

UR - http://www.scopus.com/inward/record.url?scp=85076255912&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85076255912&partnerID=8YFLogxK

U2 - 10.1145/3318216.3363316

DO - 10.1145/3318216.3363316

M3 - Conference contribution

AN - SCOPUS:85076255912

T3 - Proceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019

SP - 127

EP - 138

BT - Proceedings of the 4th ACM/IEEE Symposium on Edge Computing, SEC 2019

PB - Association for Computing Machinery, Inc

ER -