arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2404.07948cs.SEcs.AIcs.LG
收藏

面向端侧学习的嵌入式开发环境可用性与性能分析

Usability and Performance Analysis of Embedded Development Environment for On-device Learning

INSA Lyon(里昂国立应用科学学院);Inria(法国国家信息与自动化研究所);CITI(信息技术与创新中心);Université Claude Bernard Lyon 1(里昂第一大学);Ecole Centrale de Lyon(里昂中央理工学院);CNRS(法国国家科学研究中心);Ampère(安培实验室)

Enzo Scaffi, Antoine Bonneau, Frédéric Le Mouël, Fabien Mieyeville

AI总结:

本研究针对端侧TinyML应用需求,实证评估不同抽象层级的嵌入式开发工具在资源受限IoT设备上的性能与可用性,明确了Arduino Framework与RIOT OS的优劣及硬件控制层面的局限。

摘要:

本研究对可用于端侧TinyML(微型机器学习)实现的嵌入式开发工具开展实证检验。研究在资源受限的IoT(物联网)设备上,评估了不同抽象层级的各类开发工具,评估范围从基础硬件操作到极简机器学习训练部署。分析涵盖模型训练与推理过程中的内存占用、能耗、性能指标,以及不同方案的可用性。Arduino Framework(Arduino框架)实现简便,但与原生方案相比能耗更高;RIOT OS尽管内存利用率更高,却表现出高效的能耗水平,且易用性相当。操作系统中未直接集成DVFS(动态电压频率调节)等关键功能,凸显了精细硬件控制方面的局限性。

英文摘要:

This research empirically examines embedded development tools viable for on-device TinyML implementation. The research evaluates various development tools with various abstraction levels on resource-constrained IoT devices, from basic hardware manipulation to deployment of minimalistic ML training. The analysis encompasses memory usage, energy consumption, and performance metrics during model training and inference and usability of the different solutions. Arduino Framework offers ease of implementation but with increased energy consumption compared to the native option, while RIOT OS exhibits efficient energy consumption despite higher memory utilization with equivalent ease of use. The absence of certain critical functionalities like DVFS directly integrated into the OS highlights limitations for fine hardware control.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑