Adaptation of Multi-Exit Architecture for Efficient Federated Learning

Yujin Shin; Jeong Gil Ko

doi:10.1109/ICTC62082.2024.10826794

Adaptation of Multi-Exit Architecture for Efficient Federated Learning

Yujin Shin, Jeong Gil Ko

School of Integrated Technology

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

This paper explores the integration of a multi-exit architecture into federated learning (FL) to address the challenges posed by varying computational resources and non-uniform data distributions across client devices. By introducing multiple output layers, the multi-exit architecture allows clients to perform early exits based on their resource capacity, reducing computational load and communication costs. Experimental results show that this approach enhances model accuracy by up to 8.15% and reduces communication costs by 15.4%, making it a viable solution for improving FL efficiency in heterogeneous environments.

Original language	English
Title of host publication	ICTC 2024 - 15th International Conference on ICT Convergence
Subtitle of host publication	AI-Empowered Digital Innovation
Publisher	IEEE Computer Society
Pages	1740-1741
Number of pages	2
ISBN (Electronic)	9798350364637
DOIs	https://doi.org/10.1109/ICTC62082.2024.10826794
Publication status	Published - 2024
Event	15th International Conference on Information and Communication Technology Convergence, ICTC 2024 - Jeju Island, Korea, Republic of Duration: 2024 Oct 16 → 2024 Oct 18

Publication series

Name	International Conference on ICT Convergence
ISSN (Print)	2162-1233
ISSN (Electronic)	2162-1241

Conference

Conference	15th International Conference on Information and Communication Technology Convergence, ICTC 2024
Country/Territory	Korea, Republic of
City	Jeju Island
Period	24/10/16 → 24/10/18

Bibliographical note

Publisher Copyright:
© 2024 IEEE.

All Science Journal Classification (ASJC) codes

Information Systems
Computer Networks and Communications

Access to Document

10.1109/ICTC62082.2024.10826794

Cite this

@inproceedings{f4d3220a94d94aef8c9a9a9dc6e4b17e,

title = "Adaptation of Multi-Exit Architecture for Efficient Federated Learning",

abstract = "This paper explores the integration of a multi-exit architecture into federated learning (FL) to address the challenges posed by varying computational resources and non-uniform data distributions across client devices. By introducing multiple output layers, the multi-exit architecture allows clients to perform early exits based on their resource capacity, reducing computational load and communication costs. Experimental results show that this approach enhances model accuracy by up to 8.15% and reduces communication costs by 15.4%, making it a viable solution for improving FL efficiency in heterogeneous environments.",

keywords = "Federated Learning(FL), Heterogeneous Federated Learning, Mobile AI, On-device Training",

author = "Yujin Shin and Ko, {Jeong Gil}",

note = "Publisher Copyright: {\textcopyright} 2024 IEEE.; 15th International Conference on Information and Communication Technology Convergence, ICTC 2024 ; Conference date: 16-10-2024 Through 18-10-2024",

year = "2024",

doi = "10.1109/ICTC62082.2024.10826794",

language = "English",

series = "International Conference on ICT Convergence",

publisher = "IEEE Computer Society",

pages = "1740--1741",

booktitle = "ICTC 2024 - 15th International Conference on ICT Convergence",

address = "United States",

}

Shin, Y & Ko, JG 2024, Adaptation of Multi-Exit Architecture for Efficient Federated Learning. in ICTC 2024 - 15th International Conference on ICT Convergence: AI-Empowered Digital Innovation. International Conference on ICT Convergence, IEEE Computer Society, pp. 1740-1741, 15th International Conference on Information and Communication Technology Convergence, ICTC 2024, Jeju Island, Korea, Republic of, 24/10/16. https://doi.org/10.1109/ICTC62082.2024.10826794

Adaptation of Multi-Exit Architecture for Efficient Federated Learning. / Shin, Yujin; Ko, Jeong Gil.
ICTC 2024 - 15th International Conference on ICT Convergence: AI-Empowered Digital Innovation. IEEE Computer Society, 2024. p. 1740-1741 (International Conference on ICT Convergence).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Adaptation of Multi-Exit Architecture for Efficient Federated Learning

AU - Shin, Yujin

AU - Ko, Jeong Gil

PY - 2024

Y1 - 2024

N2 - This paper explores the integration of a multi-exit architecture into federated learning (FL) to address the challenges posed by varying computational resources and non-uniform data distributions across client devices. By introducing multiple output layers, the multi-exit architecture allows clients to perform early exits based on their resource capacity, reducing computational load and communication costs. Experimental results show that this approach enhances model accuracy by up to 8.15% and reduces communication costs by 15.4%, making it a viable solution for improving FL efficiency in heterogeneous environments.

AB - This paper explores the integration of a multi-exit architecture into federated learning (FL) to address the challenges posed by varying computational resources and non-uniform data distributions across client devices. By introducing multiple output layers, the multi-exit architecture allows clients to perform early exits based on their resource capacity, reducing computational load and communication costs. Experimental results show that this approach enhances model accuracy by up to 8.15% and reduces communication costs by 15.4%, making it a viable solution for improving FL efficiency in heterogeneous environments.

KW - Federated Learning(FL)

KW - Heterogeneous Federated Learning

KW - Mobile AI

KW - On-device Training

UR - http://www.scopus.com/inward/record.url?scp=85217689364&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85217689364&partnerID=8YFLogxK

U2 - 10.1109/ICTC62082.2024.10826794

DO - 10.1109/ICTC62082.2024.10826794

M3 - Conference contribution

AN - SCOPUS:85217689364

T3 - International Conference on ICT Convergence

SP - 1740

EP - 1741

BT - ICTC 2024 - 15th International Conference on ICT Convergence

PB - IEEE Computer Society

T2 - 15th International Conference on Information and Communication Technology Convergence, ICTC 2024

Y2 - 16 October 2024 through 18 October 2024

ER -

Adaptation of Multi-Exit Architecture for Efficient Federated Learning

Abstract

Publication series

Conference

Bibliographical note

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this