arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2407.05982cs.LGcs.CVcs.DC
收藏

MTL-Split:面向边缘设备的采用分割计算的多任务学习

MTL-Split: Multi-Task Learning for Edge Devices using Split Computing

University of Verona(维罗纳大学);The University of North Carolina at Chapel Hill(北卡罗来纳大学教堂山分校)

Luigi Capogrosso, Enrico Fraccaroli, Samarjit Chakraborty, Franco Fummi, Marco Cristani

AI总结:

针对边缘设备计算资源受限场景下多任务DNN在分割计算框架中的划分研究不足的问题,提出新型架构MTL-Split,在合成与真实数据上均取得良好效果。

摘要:

分割计算(Split Computing,SC)是一种新兴的有前景的方法,它将深度神经网络(Deep Neural Network,DNN)智能分割,一部分部署在边缘设备上,其余部分部署在远程服务器上。对于不允许将整个DNN远程部署、且本地没有足够计算带宽的延迟敏感型应用,该方法能够发挥DNN的算力优势。在许多此类嵌入式系统场景中(例如汽车领域的场景),计算资源受限也使得多任务学习(Multi-Task Learning,MTL)成为必要——即使用同一个DNN执行多个推理任务,而非为每个任务配备专用DNN,后者会需要更多计算带宽。然而,如何对这类多任务DNN进行划分以部署在SC框架中,目前尚未得到充分研究。本文针对这一问题展开研究,我们提出的新型架构MTL-Split在合成数据和真实世界数据上均取得了令人鼓舞的结果。源代码可在https://github.com/intelligolabs/MTL-Split获取。

英文摘要:

Split Computing (SC), where a Deep Neural Network (DNN) is intelligently split with a part of it deployed on an edge device and the rest on a remote server is emerging as a promising approach. It allows the power of DNNs to be leveraged for latency-sensitive applications that do not allow the entire DNN to be deployed remotely, while not having sufficient computation bandwidth available locally. In many such embedded systems scenarios, such as those in the automotive domain, computational resource constraints also necessitate Multi-Task Learning (MTL), where the same DNN is used for multiple inference tasks instead of having dedicated DNNs for each task, which would need more computing bandwidth. However, how to partition such a multi-tasking DNN to be deployed within a SC framework has not been sufficiently studied. This paper studies this problem, and MTL-Split, our novel proposed architecture, shows encouraging results on both synthetic and real-world data. The source code is available at https://github.com/intelligolabs/MTL-Split.

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。

↑