English

E2ATST: A Temporal-Spatial Optimized Energy-Efficient Architecture for Training Spiking Transformer

Hardware Architecture 2025-08-04 v1 Neural and Evolutionary Computing

Abstract

(1) Pengcheng Laboratory, (2) Southern University of Science and Technology, (3) Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, (4) University of Chinese Academy of Sciences

Cite

@article{arxiv.2508.00475,
  title  = {E2ATST: A Temporal-Spatial Optimized Energy-Efficient Architecture for Training Spiking Transformer},
  author = {Yunhao Ma and Yanyu Lin and Mingjing Li and Puli Quan and Chenlin Zhou and Wenyue Zhang and Zhiwei Zhong and Wanyi Jia and Xueke Zhu and Qingyan Meng and Huihui Zhou and Fengwei An},
  journal= {arXiv preprint arXiv:2508.00475},
  year   = {2025}
}
R2 v1 2026-07-01T04:29:09.917Z