Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Energy-Aware Integrated Neural Architecture Search and Partitioning for Distributed Internet of Things (IoT)

Huang, Baichuan LU orcid ; Abtahi, Azra LU and Aminifar, Amir LU orcid (2024) In IEEE Transactions on Circuits and Systems for Artificial Intelligence 1(2). p.257-271
Abstract
Internet of Things (IoT) are one of the key enablers of personalized health. However, IoT devices often have stringent constraints in terms of resources, e.g., energy budget, and, therefore, limited possibilities to exploit the state-of-the-art Deep Neural Networks (DNNs). Energy-aware Neural Architecture Search (NAS) is proposed to tackle this challenge, by exploring lightweight DNN (DNN) architectures on a single IoT device, but not leveraging the inherently distributed nature of IoT systems. As a result, the joint optimization of DNN architectures and DNN computation partitioning/offloading has not been addressed to date. In this paper, we propose an energy-aware NAS framework for distributed IoT, aiming to search for distributed Deep... (More)
Internet of Things (IoT) are one of the key enablers of personalized health. However, IoT devices often have stringent constraints in terms of resources, e.g., energy budget, and, therefore, limited possibilities to exploit the state-of-the-art Deep Neural Networks (DNNs). Energy-aware Neural Architecture Search (NAS) is proposed to tackle this challenge, by exploring lightweight DNN (DNN) architectures on a single IoT device, but not leveraging the inherently distributed nature of IoT systems. As a result, the joint optimization of DNN architectures and DNN computation partitioning/offloading has not been addressed to date. In this paper, we propose an energy-aware NAS framework for distributed IoT, aiming to search for distributed Deep Neural Networks (DNNs) to maximize prediction performance subjected to Flash Memory (Flash), Random Access Memory (RAM), and energy constraints. Our framework searches for lightweight DNN architecture with optimized prediction performance and its corresponding optimal computation partitioning to offload the partial DNN from edge to fog in a joint optimization. We evaluate our framework in the context of two common health applications, namely, seizure detection and arrhythmia classification, and demonstrate the effectiveness of our proposed joint optimization framework compared to NAS benchmarks. (Less)
Please use this url to cite or link to this publication:
author
; and
organization
publishing date
type
Contribution to journal
publication status
published
subject
in
IEEE Transactions on Circuits and Systems for Artificial Intelligence
volume
1
issue
2
pages
257 - 271
DOI
10.1109/TCASAI.2024.3493036
language
English
LU publication?
yes
id
cc8ba51d-aa29-413f-803b-44a50e6cb59f
date added to LUP
2025-03-19 15:45:58
date last changed
2025-08-12 15:55:27
@article{cc8ba51d-aa29-413f-803b-44a50e6cb59f,
  abstract     = {{Internet of Things (IoT) are one of the key enablers of personalized health. However, IoT devices often have stringent constraints in terms of resources, e.g., energy budget, and, therefore, limited possibilities to exploit the state-of-the-art Deep Neural Networks (DNNs). Energy-aware Neural Architecture Search (NAS) is proposed to tackle this challenge, by exploring lightweight DNN (DNN) architectures on a single IoT device, but not leveraging the inherently distributed nature of IoT systems. As a result, the joint optimization of DNN architectures and DNN computation partitioning/offloading has not been addressed to date. In this paper, we propose an energy-aware NAS framework for distributed IoT, aiming to search for distributed Deep Neural Networks (DNNs) to maximize prediction performance subjected to Flash Memory (Flash), Random Access Memory (RAM), and energy constraints. Our framework searches for lightweight DNN architecture with optimized prediction performance and its corresponding optimal computation partitioning to offload the partial DNN from edge to fog in a joint optimization. We evaluate our framework in the context of two common health applications, namely, seizure detection and arrhythmia classification, and demonstrate the effectiveness of our proposed joint optimization framework compared to NAS benchmarks.}},
  author       = {{Huang, Baichuan and Abtahi, Azra and Aminifar, Amir}},
  language     = {{eng}},
  month        = {{12}},
  number       = {{2}},
  pages        = {{257--271}},
  series       = {{IEEE Transactions on Circuits and Systems for Artificial Intelligence}},
  title        = {{Energy-Aware Integrated Neural Architecture Search and Partitioning for Distributed Internet of Things (IoT)}},
  url          = {{http://dx.doi.org/10.1109/TCASAI.2024.3493036}},
  doi          = {{10.1109/TCASAI.2024.3493036}},
  volume       = {{1}},
  year         = {{2024}},
}