A Comparison of Client Weighting Schemes in Federated Learning

Anna Wilbik; Barbara Pȩkala; Krzysztof Dyczkowski; Jarosław Szkoła

doi:10.1007/978-3-031-45069-3_11

A Comparison of Client Weighting Schemes in Federated Learning

Anna Wilbik, Barbara Pȩkala, Krzysztof Dyczkowski, Jarosław Szkoła

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceeding › Academic › peer-review

Abstract

Data is the new oil of the digital economy. Many business organizations are gathering and using their data to optimize their business performance. However, in some cases, an individual organization may not have a sufficient amount of data or data quality to build a well-performing model, especially in a dynamic environment. In some cases, the companies, which collect similar data, may be willing to exchange their knowledge, yet without sharing their data, e.g. due to privacy or legal issues. In such a case, federated learning is a solution. In horizontal federated learning, each client (organization) iteratively improves its model, so that it can be regularly aggregated and shared with all clients participating in the federation for further improvements. In the aggregation mechanism based on the weighted average, weights depend on the model’s quality for each client. In this paper, we extend our previous approach to federated learning with missing information, and we investigate different weighting schemes, that depend on the effectiveness of the local models. The initial results are promising.

Original language	English
Title of host publication	Uncertainty and Imprecision in Decision Making and Decision Support - New Advances, Challenges, and Perspectives
Editors	Krassimir T. Atanassov, Vassia Atanassova, Janusz Kacprzyk, Andrzej Kałuszko, Jan W. Owsiński, Eulalia Szmidt, Sławomir Zadrożny, Maciej Krawczak, Sotir S. Sotirov, Evdokia Sotirova
Publisher	Springer, Cham
Pages	116-128
Number of pages	13
ISBN (Electronic)	978-3-031-45069-3
ISBN (Print)	978-3-031-45068-6
DOIs	https://doi.org/10.1007/978-3-031-45069-3_11
Publication status	Published - 18 Oct 2023
Event	International Workshop on Intuitionistic Fuzzy Sets and Generalized Nets; National Conference on Systems and Operational Research - Warsaw, Poland Duration: 14 Oct 2022 → 14 Oct 2022 https://www2.ibspan.waw.pl/ifs2022/

Publication series

Series	Lecture Notes in Networks and Systems (LNNS)
Volume	793
ISSN	2367-3370

Conference

Conference	International Workshop on Intuitionistic Fuzzy Sets and Generalized Nets; National Conference on Systems and Operational Research
Abbreviated title	IWIFSGN 2022, BOS/SOR 2022
Country/Territory	Poland
City	Warsaw
Period	14/10/22 → 14/10/22
Internet address	https://www2.ibspan.waw.pl/ifs2022/

Access to Document

10.1007/978-3-031-45069-3_11

https://link.springer.com/10.1007/978-3-031-45069-3_11

Cite this

Wilbik, A., Pȩkala, B., Dyczkowski, K., & Szkoła, J. (2023). A Comparison of Client Weighting Schemes in Federated Learning. In K. T. Atanassov, V. Atanassova, J. Kacprzyk, A. Kałuszko, J. W. Owsiński, E. Szmidt, S. Zadrożny, M. Krawczak, S. S. Sotirov, & E. Sotirova (Eds.), Uncertainty and Imprecision in Decision Making and Decision Support - New Advances, Challenges, and Perspectives (pp. 116-128). Article Chapter 11 Springer, Cham. https://doi.org/10.1007/978-3-031-45069-3_11

Wilbik, Anna ; Pȩkala, Barbara ; Dyczkowski, Krzysztof et al. / A Comparison of Client Weighting Schemes in Federated Learning. Uncertainty and Imprecision in Decision Making and Decision Support - New Advances, Challenges, and Perspectives . editor / Krassimir T. Atanassov ; Vassia Atanassova ; Janusz Kacprzyk ; Andrzej Kałuszko ; Jan W. Owsiński ; Eulalia Szmidt ; Sławomir Zadrożny ; Maciej Krawczak ; Sotir S. Sotirov ; Evdokia Sotirova. Springer, Cham, 2023. pp. 116-128 (Lecture Notes in Networks and Systems (LNNS), Vol. 793).

@inproceedings{b56ecc3a898345b580d447b9f098c07e,

title = "A Comparison of Client Weighting Schemes in Federated Learning",

abstract = "Data is the new oil of the digital economy. Many business organizations are gathering and using their data to optimize their business performance. However, in some cases, an individual organization may not have a sufficient amount of data or data quality to build a well-performing model, especially in a dynamic environment. In some cases, the companies, which collect similar data, may be willing to exchange their knowledge, yet without sharing their data, e.g. due to privacy or legal issues. In such a case, federated learning is a solution. In horizontal federated learning, each client (organization) iteratively improves its model, so that it can be regularly aggregated and shared with all clients participating in the federation for further improvements. In the aggregation mechanism based on the weighted average, weights depend on the model{\textquoteright}s quality for each client. In this paper, we extend our previous approach to federated learning with missing information, and we investigate different weighting schemes, that depend on the effectiveness of the local models. The initial results are promising.",

author = "Anna Wilbik and Barbara Pȩkala and Krzysztof Dyczkowski and Jaros{\l}aw Szko{\l}a",

year = "2023",

month = oct,

day = "18",

doi = "10.1007/978-3-031-45069-3_11",

language = "English",

isbn = "978-3-031-45068-6",

series = "Lecture Notes in Networks and Systems (LNNS)",

publisher = "Springer, Cham",

pages = "116--128",

editor = "Atanassov, {Krassimir T.} and Vassia Atanassova and Janusz Kacprzyk and Andrzej Ka{\l}uszko and Owsi{\'n}ski, {Jan W.} and Eulalia Szmidt and S{\l}awomir Zadro{\.z}ny and Maciej Krawczak and Sotirov, {Sotir S.} and Evdokia Sotirova",

booktitle = "Uncertainty and Imprecision in Decision Making and Decision Support - New Advances, Challenges, and Perspectives",

address = "Switzerland",

note = "International Workshop on Intuitionistic Fuzzy Sets and Generalized Nets; National Conference on Systems and Operational Research, IWIFSGN 2022, BOS/SOR 2022 ; Conference date: 14-10-2022 Through 14-10-2022",

url = "https://www2.ibspan.waw.pl/ifs2022/",

}

Wilbik, A, Pȩkala, B, Dyczkowski, K & Szkoła, J 2023, A Comparison of Client Weighting Schemes in Federated Learning. in KT Atanassov, V Atanassova, J Kacprzyk, A Kałuszko, JW Owsiński, E Szmidt, S Zadrożny, M Krawczak, SS Sotirov & E Sotirova (eds), Uncertainty and Imprecision in Decision Making and Decision Support - New Advances, Challenges, and Perspectives ., Chapter 11, Springer, Cham, Lecture Notes in Networks and Systems (LNNS), vol. 793, pp. 116-128, International Workshop on Intuitionistic Fuzzy Sets and Generalized Nets; National Conference on Systems and Operational Research, Warsaw, Poland, 14/10/22. https://doi.org/10.1007/978-3-031-45069-3_11

A Comparison of Client Weighting Schemes in Federated Learning. / Wilbik, Anna; Pȩkala, Barbara; Dyczkowski, Krzysztof et al.
Uncertainty and Imprecision in Decision Making and Decision Support - New Advances, Challenges, and Perspectives . ed. / Krassimir T. Atanassov; Vassia Atanassova; Janusz Kacprzyk; Andrzej Kałuszko; Jan W. Owsiński; Eulalia Szmidt; Sławomir Zadrożny; Maciej Krawczak; Sotir S. Sotirov; Evdokia Sotirova. Springer, Cham, 2023. p. 116-128 Chapter 11 (Lecture Notes in Networks and Systems (LNNS), Vol. 793).

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceeding › Academic › peer-review

TY - GEN

T1 - A Comparison of Client Weighting Schemes in Federated Learning

AU - Wilbik, Anna

AU - Pȩkala, Barbara

AU - Dyczkowski, Krzysztof

AU - Szkoła, Jarosław

PY - 2023/10/18

Y1 - 2023/10/18

N2 - Data is the new oil of the digital economy. Many business organizations are gathering and using their data to optimize their business performance. However, in some cases, an individual organization may not have a sufficient amount of data or data quality to build a well-performing model, especially in a dynamic environment. In some cases, the companies, which collect similar data, may be willing to exchange their knowledge, yet without sharing their data, e.g. due to privacy or legal issues. In such a case, federated learning is a solution. In horizontal federated learning, each client (organization) iteratively improves its model, so that it can be regularly aggregated and shared with all clients participating in the federation for further improvements. In the aggregation mechanism based on the weighted average, weights depend on the model’s quality for each client. In this paper, we extend our previous approach to federated learning with missing information, and we investigate different weighting schemes, that depend on the effectiveness of the local models. The initial results are promising.

AB - Data is the new oil of the digital economy. Many business organizations are gathering and using their data to optimize their business performance. However, in some cases, an individual organization may not have a sufficient amount of data or data quality to build a well-performing model, especially in a dynamic environment. In some cases, the companies, which collect similar data, may be willing to exchange their knowledge, yet without sharing their data, e.g. due to privacy or legal issues. In such a case, federated learning is a solution. In horizontal federated learning, each client (organization) iteratively improves its model, so that it can be regularly aggregated and shared with all clients participating in the federation for further improvements. In the aggregation mechanism based on the weighted average, weights depend on the model’s quality for each client. In this paper, we extend our previous approach to federated learning with missing information, and we investigate different weighting schemes, that depend on the effectiveness of the local models. The initial results are promising.

U2 - 10.1007/978-3-031-45069-3_11

DO - 10.1007/978-3-031-45069-3_11

M3 - Conference article in proceeding

SN - 978-3-031-45068-6

T3 - Lecture Notes in Networks and Systems (LNNS)

SP - 116

EP - 128

BT - Uncertainty and Imprecision in Decision Making and Decision Support - New Advances, Challenges, and Perspectives

A2 - Atanassov, Krassimir T.

A2 - Atanassova, Vassia

A2 - Kacprzyk, Janusz

A2 - Kałuszko, Andrzej

A2 - Owsiński, Jan W.

A2 - Szmidt, Eulalia

A2 - Zadrożny, Sławomir

A2 - Krawczak, Maciej

A2 - Sotirov, Sotir S.

A2 - Sotirova, Evdokia

PB - Springer, Cham

T2 - International Workshop on Intuitionistic Fuzzy Sets and Generalized Nets; National Conference on Systems and Operational Research

Y2 - 14 October 2022 through 14 October 2022

ER -

Wilbik A, Pȩkala B, Dyczkowski K, Szkoła J. A Comparison of Client Weighting Schemes in Federated Learning. In Atanassov KT, Atanassova V, Kacprzyk J, Kałuszko A, Owsiński JW, Szmidt E, Zadrożny S, Krawczak M, Sotirov SS, Sotirova E, editors, Uncertainty and Imprecision in Decision Making and Decision Support - New Advances, Challenges, and Perspectives . Springer, Cham. 2023. p. 116-128. Chapter 11. (Lecture Notes in Networks and Systems (LNNS), Vol. 793). doi: 10.1007/978-3-031-45069-3_11