Presentation 2023-08-31
Performance Analysis of On-device Hierarchical Federated Learning Frameworks
Zhaoyang Du, Celimuge Wu, Tsutomu Yoshinaga,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In the rapidly advancing field of artificial intelligence and deep learning, centralized architectures exhibit inherent limitations such as high latency and computational overhead. This study introduces a Hierarchical Federated Learning (HFL) framework to overcome these challenges by decentralizing the learning process. Employing intermediary edge servers, HFL efficiently distributes computational and communication loads, as demonstrated through experiments with Raspberry Pi devices and the Cifar10 dataset. The results reveal significant improvements in model accuracy and convergence speed over traditional cloud-based Federated Learning (FL), showcasing the potential of HFL to enhance FL performance with acceptable communication costs. This innovation marks a step towards more resilient and adaptive learning systems in areas such as Intelligent Transportation Systems (ITS) and Internet of Things (IoT) devices.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Hierarchical Federated Learning / Edge Computing / IoT / Intelligent Transportation Systems
Paper # CQ2023-28
Date of Issue 2023-08-24 (CQ)

Conference Information
Committee CQ
Conference Date 2023/8/31(2days)
Place (in Japanese) (See Japanese page)
Place (in English) Tenjin-Misaki Sports Park
Topics (in Japanese) (See Japanese page)
Topics (in English) Wireless Communications Quality, 6G, IoT, Resource Management, Wireless Transmission, Cross layer Technologies, etc.
Chair Takefumi Hiraguri(Nippon Inst. of Tech.)
Vice Chair Takahiro Matsuda(Tokyo Metropolitan Univ.) / Gou Hasegawa(Tohoku Univ.) / Sumaru Niida(KDDI Research)
Secretary Takahiro Matsuda(NTT) / Gou Hasegawa(Tama Univ.) / Sumaru Niida(Tsukuba Univ.)
Assistant Ryo Nakamura(Fukuoka Univ.) / Toshiro Nakahira(NTT) / Kenta Tsukatsune(Okayama Univ. of Science)

Paper Information
Registration To Technical Committee on Communication Quality
Language ENG
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Performance Analysis of On-device Hierarchical Federated Learning Frameworks
Sub Title (in English)
Keyword(1) Hierarchical Federated Learning
Keyword(2) Edge Computing
Keyword(3) IoT
Keyword(4) Intelligent Transportation Systems
1st Author's Name Zhaoyang Du
1st Author's Affiliation The University of Electro-Communications(UEC)
2nd Author's Name Celimuge Wu
2nd Author's Affiliation The University of Electro-Communications(UEC)
3rd Author's Name Tsutomu Yoshinaga
3rd Author's Affiliation The University of Electro-Communications(UEC)
Date 2023-08-31
Paper # CQ2023-28
Volume (vol) vol.123
Number (no) CQ-174
Page pp.pp.14-19(CQ),
#Pages 6
Date of Issue 2023-08-24 (CQ)