EUVIMEDEUVIMED entdecken
Crossref

Feature Overlapping: Temporal Differential Decoupling for Efficient Spiking Neural Network Training

Yuqian Liu iD, Yuechao Wang, Yizhou Jiang, Guanyu Chen, Tianrun Xu, Feng Chen

Annals of the New York Academy of Sciences · 2026

DOI: 10.1111/nyas.70204 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Annals of the New York Academy of Sciences, 2026. Verfasst von Yuqian Liu und 5 weiteren Autor:innen. Beteiligte Einrichtung: Department of Automation, Tsinghua University, Beijing, China. Hauptthemen laut MEDLINE: Neural Networks, Computer, Action Potentials, Neurons, Models, Neurological. Die Arbeit nennt 50 Literaturangaben. Gefördert von: National Key Research and Development Program of China under STI 2030-Major Projects, National Key Research and Development Program of China, National Natural Science Foundation of China.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Spiking neural networks (SNNs) possess great potential for energy‐efficient computation; however, their practical deployment is often constrained by the high computational cost associated with training across multiple time steps. Unlike previous approaches that simply reduce the number of time steps, this work investigates temporal feature redundancy and reveals the feature overlapping phenomenon in SNNs—showing that substantial computational redundancy exists across temporal dimensions. To address this issue, we propose our core contribution, temporal differential decoupling (TDD), which transforms network computation into the differential domain to disentangle static and dynamic feature components. This decoupling enables focused processing of informative signals and significantly reduces redundant computation without compromising model accuracy. Building upon this framework, we further design the TDD‐based differential domain low‐sparsity approximation (TDD‐DDLA) algorithm, which is based on the gradient sensitivity criterion, as an implementation strategy to quantify each temporal feature's contribution to gradient updates and achieve efficient energy optimization. In contrast to prior studies that only adjusted time steps without explicit feature‐level analysis, our framework provides a structured and theoretically grounded analysis of temporal feature evolution. Experimental results demonstrate that our method achieves up to 80.9% fewer spikes per time step and 57.8% fewer total spikes, with no degradation in classification performance, offering a promising pathway toward scalable, low‐cost, and high‐accuracy SNN deployment.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Neural Networks, ComputerAction PotentialsNeuronsModels, NeurologicalNerve NetAlgorithmsAnimalsHumansSoft Computing

Stichwörter der Autor:innen

feature analysisimage classificationspiking neural networks (SNNs)transformer

Autor:innen und Institutionen

  • Department of Automation, Tsinghua University, Beijing, China.

ORCID-Profile: Yuqian Liu

Förderung

  • National Key Research and Development Program of China under STI 2030-Major Projects
  • National Key Research and Development Program of China
  • National Natural Science Foundation of China

Zitieren

APA

Liu, Y., Wang, Y., Jiang, Y., Chen, G., Xu, T., & Chen, F. (2026). Feature Overlapping: Temporal Differential Decoupling for Efficient Spiking Neural Network Training. Annals of the New York Academy of Sciences. https://doi.org/10.1111/nyas.70204

Vancouver

Liu Y, Wang Y, Jiang Y, Chen G, Xu T, Chen F. Feature Overlapping: Temporal Differential Decoupling for Efficient Spiking Neural Network Training. Annals of the New York Academy of Sciences. 2026. doi:10.1111/nyas.70204

BibTeX
@article{Liu2026,
  author = {Liu, Yuqian and Wang, Yuechao and Jiang, Yizhou and Chen, Guanyu and Xu, Tianrun and Chen, Feng},
  title = {Feature Overlapping: Temporal Differential Decoupling for Efficient Spiking Neural Network Training},
  journal = {Annals of the New York Academy of Sciences},
  year = {2026},
  doi = {10.1111/nyas.70204}
}
Kostenlos starten →