AITOP

Spiking Transformers

共 1 条相关 AI 资讯
8月24日
09:58
09:58官方账号arXiv cs.LG@Tengteng Lei, Prabodh Katti, Rashi Dutt, Houssem Sifaou, Tan Peng, Osvaldo Simeone, Kai Xu, Bipin Rajendran
Proposes an implicit-perturbation ZO (IPZO) architecture for fine-tuning spiking neural networks, reducing RMW operations and hardware footprint. PGU-XOR achieves similar accuracy to software RNGs with lower energy overhead. IPZO reduces perturbation energy significantly compared to explicit weight perturbation.
论文Spiking TransformersZeroth-Order OptimizationEvent-Triggered Perturbation
推荐理由：This paper introduces a novel approach for fine-tuning spiking neural networks with reduced energy consumption and improved accuracy, offering a significant advancement in the field of event-driven computing.
原文