ApDeepSense: Deep learning uncertainty estimation without the pain for IoT applications

Shuochao Yao, Yiran Zhao, Huajie Shao, Chao Zhang, Aston Zhang, Dongxin Liu, Shengzhong Liu, Lu Su, Tarek Abdelzaher

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

Abstract

Recent advances in deep-learning-based applications have attracted a growing attention from the IoT community. These highly capable learning models have shown significant improvements in expected accuracy of various sensory inference tasks. One important and yet overlooked direction remains to provide uncertainty estimates in deep learning outputs. Since robustness and reliability of sensory inference results are critical to IoT systems, uncertainty estimates are indispensable for IoT applications. To address this challenge, we develop ApDeepSense, an effective and efficient deep learning uncertainty estimation method for resource-constrained IoT devices. ApDeepSense leverages an implicit Bayesian approximation that links neural networks to deep Gaussian processes, allowing output uncertainty to be quantified. Our approach is shown to significantly reduce the execution time and energy consumption of uncertainty estimation thanks to a novel layer-wise approximation that replaces the traditional computationally intensive sampling-based uncertainty estimation methods. ApDeepSense is designed for neural net-works trained using dropout; one of the most widely used regularization methods in deep learning. No additional training is needed for uncertainty estimation purposes. We evaluate ApDeepSense using four IoT applications on Intel Edison devices. Results show that ApDeepSense can reduce around 88.9% of the execution time and 90.0% of the energy consumption, while producing more accurate uncertainty estimates compared with state-of-the-art methods.

Original languageEnglish (US)
Title of host publicationProceedings - 2018 IEEE 38th International Conference on Distributed Computing Systems, ICDCS 2018
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages334-343
Number of pages10
ISBN (Electronic)9781538668719
DOIs
StatePublished - Jul 19 2018
Event38th IEEE International Conference on Distributed Computing Systems, ICDCS 2018 - Vienna, Austria
Duration: Jul 2 2018Jul 5 2018

Publication series

NameProceedings - International Conference on Distributed Computing Systems
Volume2018-July

Other

Other38th IEEE International Conference on Distributed Computing Systems, ICDCS 2018
CountryAustria
CityVienna
Period7/2/187/5/18

Fingerprint

Energy utilization
Neural networks
Deep learning
Uncertainty
Internet of things
Sampling

Keywords

  • Deep learning
  • Internet of Things
  • Mobile Computing
  • Uncertainty estimation

ASJC Scopus subject areas

  • Software
  • Hardware and Architecture
  • Computer Networks and Communications

Cite this

Yao, S., Zhao, Y., Shao, H., Zhang, C., Zhang, A., Liu, D., ... Abdelzaher, T. (2018). ApDeepSense: Deep learning uncertainty estimation without the pain for IoT applications. In Proceedings - 2018 IEEE 38th International Conference on Distributed Computing Systems, ICDCS 2018 (pp. 334-343). (Proceedings - International Conference on Distributed Computing Systems; Vol. 2018-July). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICDCS.2018.00041

ApDeepSense : Deep learning uncertainty estimation without the pain for IoT applications. / Yao, Shuochao; Zhao, Yiran; Shao, Huajie; Zhang, Chao; Zhang, Aston; Liu, Dongxin; Liu, Shengzhong; Su, Lu; Abdelzaher, Tarek.

Proceedings - 2018 IEEE 38th International Conference on Distributed Computing Systems, ICDCS 2018. Institute of Electrical and Electronics Engineers Inc., 2018. p. 334-343 (Proceedings - International Conference on Distributed Computing Systems; Vol. 2018-July).

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

Yao, S, Zhao, Y, Shao, H, Zhang, C, Zhang, A, Liu, D, Liu, S, Su, L & Abdelzaher, T 2018, ApDeepSense: Deep learning uncertainty estimation without the pain for IoT applications. in Proceedings - 2018 IEEE 38th International Conference on Distributed Computing Systems, ICDCS 2018. Proceedings - International Conference on Distributed Computing Systems, vol. 2018-July, Institute of Electrical and Electronics Engineers Inc., pp. 334-343, 38th IEEE International Conference on Distributed Computing Systems, ICDCS 2018, Vienna, Austria, 7/2/18. https://doi.org/10.1109/ICDCS.2018.00041
Yao S, Zhao Y, Shao H, Zhang C, Zhang A, Liu D et al. ApDeepSense: Deep learning uncertainty estimation without the pain for IoT applications. In Proceedings - 2018 IEEE 38th International Conference on Distributed Computing Systems, ICDCS 2018. Institute of Electrical and Electronics Engineers Inc. 2018. p. 334-343. (Proceedings - International Conference on Distributed Computing Systems). https://doi.org/10.1109/ICDCS.2018.00041
Yao, Shuochao ; Zhao, Yiran ; Shao, Huajie ; Zhang, Chao ; Zhang, Aston ; Liu, Dongxin ; Liu, Shengzhong ; Su, Lu ; Abdelzaher, Tarek. / ApDeepSense : Deep learning uncertainty estimation without the pain for IoT applications. Proceedings - 2018 IEEE 38th International Conference on Distributed Computing Systems, ICDCS 2018. Institute of Electrical and Electronics Engineers Inc., 2018. pp. 334-343 (Proceedings - International Conference on Distributed Computing Systems).
@inproceedings{b356bda2c7ed46ba9a8ecc7722dd178f,
title = "ApDeepSense: Deep learning uncertainty estimation without the pain for IoT applications",
abstract = "Recent advances in deep-learning-based applications have attracted a growing attention from the IoT community. These highly capable learning models have shown significant improvements in expected accuracy of various sensory inference tasks. One important and yet overlooked direction remains to provide uncertainty estimates in deep learning outputs. Since robustness and reliability of sensory inference results are critical to IoT systems, uncertainty estimates are indispensable for IoT applications. To address this challenge, we develop ApDeepSense, an effective and efficient deep learning uncertainty estimation method for resource-constrained IoT devices. ApDeepSense leverages an implicit Bayesian approximation that links neural networks to deep Gaussian processes, allowing output uncertainty to be quantified. Our approach is shown to significantly reduce the execution time and energy consumption of uncertainty estimation thanks to a novel layer-wise approximation that replaces the traditional computationally intensive sampling-based uncertainty estimation methods. ApDeepSense is designed for neural net-works trained using dropout; one of the most widely used regularization methods in deep learning. No additional training is needed for uncertainty estimation purposes. We evaluate ApDeepSense using four IoT applications on Intel Edison devices. Results show that ApDeepSense can reduce around 88.9{\%} of the execution time and 90.0{\%} of the energy consumption, while producing more accurate uncertainty estimates compared with state-of-the-art methods.",
keywords = "Deep learning, Internet of Things, Mobile Computing, Uncertainty estimation",
author = "Shuochao Yao and Yiran Zhao and Huajie Shao and Chao Zhang and Aston Zhang and Dongxin Liu and Shengzhong Liu and Lu Su and Tarek Abdelzaher",
year = "2018",
month = "7",
day = "19",
doi = "10.1109/ICDCS.2018.00041",
language = "English (US)",
series = "Proceedings - International Conference on Distributed Computing Systems",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "334--343",
booktitle = "Proceedings - 2018 IEEE 38th International Conference on Distributed Computing Systems, ICDCS 2018",
address = "United States",

}

TY - GEN

T1 - ApDeepSense

T2 - Deep learning uncertainty estimation without the pain for IoT applications

AU - Yao, Shuochao

AU - Zhao, Yiran

AU - Shao, Huajie

AU - Zhang, Chao

AU - Zhang, Aston

AU - Liu, Dongxin

AU - Liu, Shengzhong

AU - Su, Lu

AU - Abdelzaher, Tarek

PY - 2018/7/19

Y1 - 2018/7/19

N2 - Recent advances in deep-learning-based applications have attracted a growing attention from the IoT community. These highly capable learning models have shown significant improvements in expected accuracy of various sensory inference tasks. One important and yet overlooked direction remains to provide uncertainty estimates in deep learning outputs. Since robustness and reliability of sensory inference results are critical to IoT systems, uncertainty estimates are indispensable for IoT applications. To address this challenge, we develop ApDeepSense, an effective and efficient deep learning uncertainty estimation method for resource-constrained IoT devices. ApDeepSense leverages an implicit Bayesian approximation that links neural networks to deep Gaussian processes, allowing output uncertainty to be quantified. Our approach is shown to significantly reduce the execution time and energy consumption of uncertainty estimation thanks to a novel layer-wise approximation that replaces the traditional computationally intensive sampling-based uncertainty estimation methods. ApDeepSense is designed for neural net-works trained using dropout; one of the most widely used regularization methods in deep learning. No additional training is needed for uncertainty estimation purposes. We evaluate ApDeepSense using four IoT applications on Intel Edison devices. Results show that ApDeepSense can reduce around 88.9% of the execution time and 90.0% of the energy consumption, while producing more accurate uncertainty estimates compared with state-of-the-art methods.

AB - Recent advances in deep-learning-based applications have attracted a growing attention from the IoT community. These highly capable learning models have shown significant improvements in expected accuracy of various sensory inference tasks. One important and yet overlooked direction remains to provide uncertainty estimates in deep learning outputs. Since robustness and reliability of sensory inference results are critical to IoT systems, uncertainty estimates are indispensable for IoT applications. To address this challenge, we develop ApDeepSense, an effective and efficient deep learning uncertainty estimation method for resource-constrained IoT devices. ApDeepSense leverages an implicit Bayesian approximation that links neural networks to deep Gaussian processes, allowing output uncertainty to be quantified. Our approach is shown to significantly reduce the execution time and energy consumption of uncertainty estimation thanks to a novel layer-wise approximation that replaces the traditional computationally intensive sampling-based uncertainty estimation methods. ApDeepSense is designed for neural net-works trained using dropout; one of the most widely used regularization methods in deep learning. No additional training is needed for uncertainty estimation purposes. We evaluate ApDeepSense using four IoT applications on Intel Edison devices. Results show that ApDeepSense can reduce around 88.9% of the execution time and 90.0% of the energy consumption, while producing more accurate uncertainty estimates compared with state-of-the-art methods.

KW - Deep learning

KW - Internet of Things

KW - Mobile Computing

KW - Uncertainty estimation

UR - http://www.scopus.com/inward/record.url?scp=85050967144&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85050967144&partnerID=8YFLogxK

U2 - 10.1109/ICDCS.2018.00041

DO - 10.1109/ICDCS.2018.00041

M3 - Conference contribution

AN - SCOPUS:85050967144

T3 - Proceedings - International Conference on Distributed Computing Systems

SP - 334

EP - 343

BT - Proceedings - 2018 IEEE 38th International Conference on Distributed Computing Systems, ICDCS 2018

PB - Institute of Electrical and Electronics Engineers Inc.

ER -