Publication Details

Evolutionary Neural Architecture Search Supporting Approximate Multipliers

PIŇOS Michal, MRÁZEK Vojtěch and SEKANINA Lukáš. Evolutionary Neural Architecture Search Supporting Approximate Multipliers. In: Genetic Programming, 24th European Conference, EuroGP 2021. Lecture Notes in Computer Science, vol 12691, vol. 12691. Seville: Springer Nature Switzerland AG, 2021, pp. 82-97. ISBN 978-3-030-72812-0. Available from: https://link.springer.com/chapter/10.1007%2F978-3-030-72812-0_6
Czech title
Evoluční návrh architektur neuronových sítí podporující aproximativní násobičky
Type
conference paper
Language
english
Authors
URL
Keywords

Approximate computing, Convolutional neural network, Cartesian genetic programming, Neuroevolution, Energy efficiency 

Abstract

There is a growing interest in automated neural architecture search (NAS) methods. They are employed to routinely deliver high-quality neural network architectures for various challenging data sets and reduce the designer's effort. The NAS methods utilizing multi-objective evolutionary algorithms are especially useful when the objective is not only to minimize the network error but also to minimize the number of parameters (weights) or power consumption of the inference phase. We propose a multi-objective NAS method based on Cartesian genetic programming for evolving convolutional neural networks (CNN). The method allows approximate operations to be used in CNNs to reduce the power consumption of a target hardware implementation. During the NAS process, a suitable CNN architecture is evolved together with approximate multipliers to deliver the best trade-offs between the accuracy, network size, and power consumption. The most suitable approximate multipliers are automatically selected from a library of approximate multipliers. Evolved CNNs are compared with common human-created CNNs of a similar complexity on the CIFAR-10 benchmark problem.

Published
2021
Pages
82-97
Proceedings
Genetic Programming, 24th European Conference, EuroGP 2021
Series
Lecture Notes in Computer Science, vol 12691
Volume
12691
Conference
24th European Conference on Genetic Programming, Seville, ES
ISBN
978-3-030-72812-0
Publisher
Springer Nature Switzerland AG
Place
Seville, ES
DOI
UT WoS
000894232700006
EID Scopus
BibTeX
@INPROCEEDINGS{FITPUB12361,
   author = "Michal Pi\v{n}os and Vojt\v{e}ch Mr\'{a}zek and Luk\'{a}\v{s} Sekanina",
   title = "Evolutionary Neural Architecture Search Supporting Approximate Multipliers",
   pages = "82--97",
   booktitle = "Genetic Programming, 24th European Conference, EuroGP 2021",
   series = "Lecture Notes in Computer Science, vol 12691",
   volume = 12691,
   year = 2021,
   location = "Seville, ES",
   publisher = "Springer Nature Switzerland AG",
   ISBN = "978-3-030-72812-0",
   doi = "10.1007/978-3-030-72812-0\_6",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/12361"
}
Back to top