EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

An Efficient Evolutionary Neural Architecture Search Algorithm Without Training

Yang An, Changsheng Zhang iD, Jintao Shao, Yuxiao Yan, Baiqing Sun

Biomimetics · 29.06.2025 · 10 (7) · 421

DOI: 10.3390/biomimetics10070421 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Biomimetics, online erschienen am 29.06.2025. Verfasst von Yang An und 4 weiteren Autor:innen. Beteiligte Einrichtung: Software College, Northeastern University, Shenyang 110167, China. Die Arbeit wurde bisher 11-mal zitiert (Crossref) und nennt 80 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: the Key Project of National Natural Science Foundation of China, the Local Science and Technology Development Fund Project Under the Guidance of the Central Government of China, the Taking Lead Science and Technology Research Project of Liaoning.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Neural Architecture Search (NAS) has made significant advancements in autonomously constructing high-performance network architectures, capturing extensive attention. However, a key challenge of existing NAS approaches is the intensive performance evaluation, leading to significant time and computational resource consumption. In this paper, we propose an efficient Evolutionary Neural Architecture Search (ENAS) method to address this issue. Specifically, in order to accelerate the convergence speed of the algorithm and shorten the search time, thereby avoiding blind searching in the early stages of the algorithm, we drew on the principles of biometrics to redesign the interaction between individuals in the evolutionary algorithm. By making full use of the information carried by individuals, we promoted information exchange and optimization between individuals and their neighbors, thereby improving local search capabilities while maintaining global search capabilities. Furthermore, to accelerate the evaluation process and minimize computational resource consumption, a multi-metric training-free evaluator is introduced to assess network performance, bypassing the resource-intensive training phase, and the adopted multi-metric combination method further solves the ranking offset problem. To evaluate the performance of the proposed method, we conduct experiments on two widely adopted benchmarks, NAS-Bench-101 and NAS-Bench-201. Comparative analysis with state-of-the-art algorithms shows that our proposed method identifies network architectures with comparable or better performance while requiring significantly less time.

Schlagwörter

Stichwörter der Autor:innen

evolutionary algorithmindividual interactionneural architecture searchtraining-free

Autor:innen und Institutionen

  • Software College, Northeastern University, Shenyang 110167, China.
  • College of Computer Science and Engineering, Ningxia Institute of Science and Technology, Shizuishan 753000, China.
  • Lightspeed Studios, Tencent Co., Ltd., Shenzhen 518000, China.

ORCID-Profile: Changsheng Zhang

Förderung

  • the Key Project of National Natural Science Foundation of China
  • the Local Science and Technology Development Fund Project Under the Guidance of the Central Government of China
  • the Taking Lead Science and Technology Research Project of Liaoning
  • National Natural Science Foundation of China

Zitieren

APA

An, Y., Zhang, C., Shao, J., Yan, Y., & Sun, B. (2025). An Efficient Evolutionary Neural Architecture Search Algorithm Without Training. Biomimetics, 10(7), 421. https://doi.org/10.3390/biomimetics10070421

Vancouver

An Y, Zhang C, Shao J, Yan Y, Sun B. An Efficient Evolutionary Neural Architecture Search Algorithm Without Training. Biomimetics. 2025;10(7):421. doi:10.3390/biomimetics10070421

BibTeX
@article{An2025,
  author = {An, Yang and Zhang, Changsheng and Shao, Jintao and Yan, Yuxiao and Sun, Baiqing},
  title = {An Efficient Evolutionary Neural Architecture Search Algorithm Without Training},
  journal = {Biomimetics},
  year = {2025},
  volume = {10},
  number = {7},
  pages = {421},
  doi = {10.3390/biomimetics10070421}
}
Kostenlos starten →