Wireless Channel Adaptive DNN Split Inference for Resource-Constrained Edge Devices

被引：6

作者：

Lee, Jaeduk ^{[1
,2
]}

Lee, Hojung ^{[3
]}

Choi, Wan ^{[1
,2
]}

机构：

[1] Seoul Natl Univ SNU, Inst New Media & Commun, Seoul 08826, South Korea

[2] Seoul Natl Univ SNU, Dept Elect & Comp Engn, Seoul 08826, South Korea

[3] Korea Adv Inst Sci & Technol KAIST, Sch Elect Engn, Daejeon 34141, South Korea

来源：

IEEE COMMUNICATIONS LETTERS | 2023年 / 27卷 / 06期

基金：

新加坡国家研究基金会;

关键词：

Servers; Wireless communication; Performance evaluation; Uplink; Energy consumption; Downlink; Memory management; Deep learning; split inference; wireless channels; INTELLIGENCE;

D O I：

10.1109/LCOMM.2023.3269769

中图分类号：

TN [电子技术、通信技术];

学科分类号：

0809 ;

摘要：

Split inference facilitates deep neural network (DNN) inference tasks at resource-constrained edge devices. However, a pre-determined split configuration of a DNN limits the inference performance in time-varying wireless channels. To accelerate the inference, we propose a two-stage wireless channel adaptive split inference method by considering memory and energy constraints on the edge device. The proposed scheme is able to offer the privacy of the edge device and improves inference performance in time-varying wireless channels by leveraging a U-shaped DNN splitting framework and adaptively determining the splitting points of a DNN in real-time according to time-varying wireless channel gains.

引用

页码：1520 / 1524

页数：5