Aarhus Universitets segl

Qi Zhang

Receptive Field-based Segmentation for Distributed CNN Inference Acceleration in Collaborative Edge Computing

Publikation: Bidrag til bog/antologi/rapport/proceedingKonferencebidrag i proceedingsForskningpeer review

This paper studies inference acceleration using distributed convolutional neural networks (CNNs) in collaborative edge computing network. To avoid inference accuracy loss in inference task partitioning, we propose receptive field-based segmentation (RFS). To reduce the computation time and communication overhead, we propose a novel collaborative edge computing using fused-layer parallelization to partition a CNN model into multiple blocks of convolutional layers. In this scheme, the collaborative edge servers (ESs) only need to exchange small fraction of the sub-outputs after computing each fused block. In addition, to find the optimal solution of partitioning a CNN model into multiple blocks, we use dynamic programming, named as dynamic programming for fused-layer parallelization (DPFP). The experimental results show that DPFP can accelerate inference of VGG-16 up to 73% compared with the pre-trained model, which outperforms the existing work MoDNN in all tested scenarios. Moreover, we evaluate the service reliability of DPFP under time-variant channel, which shows that DPFP is an effective solution to ensure high service reliability with strict service deadline.

TitelICC 2022 - IEEE International Conference on Communications
Antal sider6
ISBN (trykt)978-1-5386-8348-4
ISBN (Elektronisk)978-1-5386-8347-7
StatusUdgivet - 2022
BegivenhedIEEE International Conference on Communications - Coex, Seoul, Sydkorea
Varighed: 16 maj 202220 maj 2022


KonferenceIEEE International Conference on Communications

Bibliografisk note

Funding Information:
ACKNOWLEDGMENT This work is supported by Agile-IoT project (Grant No. 9131-00119B) granted by the Danish Council for Independent Research.

Publisher Copyright:
© 2022 IEEE.

Se relationer på Aarhus Universitet Citationsformater

ID: 278390204