Academic Bibliography

 Search 200 years of publications by Ghent University researchers.
Advanced search
1 file | 2.21 MB
Add to list
  1. Search results
  2. Iterative neural networks for adaptiv...

Iterative neural networks for adaptive inference on resource-constrained devices

Sam Leroux (UGent) , Tim Verbelen (UGent) , Pieter Simoens (UGent) and Bart Dhoedt (UGent)
(2022) NEURAL COMPUTING & APPLICATIONS. 34(13). p.10321-10336
Author
Organization
Project
Abstract
The computational cost of evaluating a neural network usually only depends on design choices such as the number of layers or the number of units in each layer and not on the actual input. In this work, we build upon deep Residual Networks (ResNets) and use their properties to design a more efficient adaptive neural network building block. We propose a new architecture, which replaces the sequential layers with an iterative structure where weights are reused multiple times for a single input image, reducing the storage requirements drastically. In addition, we incorporate an adaptive computation module that allows the network to adjust its computational cost at run time for each input sample independently. We experimentally validate our models on image classification, object detection and semantic segmentation tasks and show that our models only use their full capacity for the hardest input samples and are more efficient on average.
Keywords
INTERNET, TERM, Efficient deep neural networks, Inference on the edge, Adaptive, computation, Resource-constrained deep learning

Downloads

  • Download
    (...).pdf
    • full text (Published version)
    • |
    • UGent only
    • |
    • PDF
    • |
    • 2.21 MB

Citation

Please use this url to cite or link to this publication:

MLA
Leroux, Sam, et al. “Iterative Neural Networks for Adaptive Inference on Resource-Constrained Devices.” NEURAL COMPUTING & APPLICATIONS, vol. 34, no. 13, 2022, pp. 10321–36, doi:10.1007/s00521-022-06910-5.
APA
Leroux, S., Verbelen, T., Simoens, P., & Dhoedt, B. (2022). Iterative neural networks for adaptive inference on resource-constrained devices. NEURAL COMPUTING & APPLICATIONS, 34(13), 10321–10336. https://doi.org/10.1007/s00521-022-06910-5
Chicago author-date
Leroux, Sam, Tim Verbelen, Pieter Simoens, and Bart Dhoedt. 2022. “Iterative Neural Networks for Adaptive Inference on Resource-Constrained Devices.” NEURAL COMPUTING & APPLICATIONS 34 (13): 10321–36. https://doi.org/10.1007/s00521-022-06910-5.
Chicago author-date (all authors)
Leroux, Sam, Tim Verbelen, Pieter Simoens, and Bart Dhoedt. 2022. “Iterative Neural Networks for Adaptive Inference on Resource-Constrained Devices.” NEURAL COMPUTING & APPLICATIONS 34 (13): 10321–10336. doi:10.1007/s00521-022-06910-5.
Vancouver
1.
Leroux S, Verbelen T, Simoens P, Dhoedt B. Iterative neural networks for adaptive inference on resource-constrained devices. NEURAL COMPUTING & APPLICATIONS. 2022;34(13):10321–36.
IEEE
[1]
S. Leroux, T. Verbelen, P. Simoens, and B. Dhoedt, “Iterative neural networks for adaptive inference on resource-constrained devices,” NEURAL COMPUTING & APPLICATIONS, vol. 34, no. 13, pp. 10321–10336, 2022.
@article{8742510,
  abstract     = {{The computational cost of evaluating a neural network usually only depends on design choices such as the number of layers or the number of units in each layer and not on the actual input. In this work, we build upon deep Residual Networks (ResNets) and use their properties to design a more efficient adaptive neural network building block. We propose a new architecture, which replaces the sequential layers with an iterative structure where weights are reused multiple times for a single input image, reducing the storage requirements drastically. In addition, we incorporate an adaptive computation module that allows the network to adjust its computational cost at run time for each input sample independently. We experimentally validate our models on image classification, object detection and semantic segmentation tasks and show that our models only use their full capacity for the hardest input samples and are more efficient on average.}},
  author       = {{Leroux, Sam and Verbelen, Tim and Simoens, Pieter and Dhoedt, Bart}},
  issn         = {{0941-0643}},
  journal      = {{NEURAL COMPUTING & APPLICATIONS}},
  keywords     = {{INTERNET,TERM,Efficient deep neural networks,Inference on the edge,Adaptive,computation,Resource-constrained deep learning}},
  language     = {{eng}},
  number       = {{13}},
  pages        = {{10321--10336}},
  title        = {{Iterative neural networks for adaptive inference on resource-constrained devices}},
  url          = {{http://doi.org/10.1007/s00521-022-06910-5}},
  volume       = {{34}},
  year         = {{2022}},
}
Altmetric
View in Altmetric
Web of Science
Times cited: