Sparse Incremental Aggregation in Satellite Federated Learning

Nasrin Razmi; Sourav Mukherjee; Bho Matthiesen; Armin Dekorsy; Petar Popovski

doi:10.1109/IEEECONF62907.2025.10949091

Sparse Incremental Aggregation in Satellite Federated Learning

Nasrin Razmi, Sourav Mukherjee, Bho Matthiesen, Armin Dekorsy, Petar Popovski

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

Abstract

This paper studies Federated Learning (FL) in low Earth orbit (LEO) satellite constellations, where satellites are connected via intra-orbit inter-satellite links (ISLs) to their neighboring satellites. During the FL training process, satellites in each orbit forward gradients from nearby satellites, which are eventually transferred to the parameter server (PS). To enhance the efficiency of the FL training process, satellites apply in-network aggregation, referred to as incremental aggregation. In this work, the gradient sparsification methods from [1] are applied to satellite scenarios to improve bandwidth efficiency during incremental aggregation. The numerical results highlight an increase of over 4 x in bandwidth efficiency as the number of satellite in the orbital plane increases.

Original language	English
Title of host publication	2025 14th International ITG Conference on Systems, Communications and Coding (SCC)
Number of pages	6
Publisher	IEEE (Institute of Electrical and Electronics Engineers)
Publication date	13 Mar 2025
Pages	1-6
Article number	10949091
ISBN (Print)	979-8-3315-2290-2
ISBN (Electronic)	9798331522896
DOIs	https://doi.org/10.1109/IEEECONF62907.2025.10949091
Publication status	Published - 13 Mar 2025
Event	2025 14th International ITG Conference on Systems, Communications and Coding (SCC) - Karlsruhe, Germany Duration: 10 Mar 2025 → 13 Mar 2025

Conference

Conference	2025 14th International ITG Conference on Systems, Communications and Coding (SCC)
Location	Karlsruhe, Germany
Period	10/03/2025 → 13/03/2025

Keywords

Federated learning
Load modeling
Low earth orbit satellites
Numerical models
Orbits
Satellite constellations
Satellites
Servers
Spectral efficiency
Training
Satellite Constellation
gradient sparsification
in-network computing

Access to Document

10.1109/IEEECONF62907.2025.10949091

https://arxiv.org/abs/2501.11385v1Licence: CC BY 4.0

AUB Link

Search for the material in Aalborg University Library's search engine

Cite this

@inproceedings{b65ecd47d450456c9319822d694c7d2d,

title = "Sparse Incremental Aggregation in Satellite Federated Learning",

abstract = "This paper studies Federated Learning (FL) in low Earth orbit (LEO) satellite constellations, where satellites are connected via intra-orbit inter-satellite links (ISLs) to their neighboring satellites. During the FL training process, satellites in each orbit forward gradients from nearby satellites, which are eventually transferred to the parameter server (PS). To enhance the efficiency of the FL training process, satellites apply in-network aggregation, referred to as incremental aggregation. In this work, the gradient sparsification methods from [1] are applied to satellite scenarios to improve bandwidth efficiency during incremental aggregation. The numerical results highlight an increase of over 4 x in bandwidth efficiency as the number of satellite in the orbital plane increases.",

keywords = "Federated learning, Load modeling, Low earth orbit satellites, Numerical models, Orbits, Satellite constellations, Satellites, Servers, Spectral efficiency, Training, Satellite Constellation, gradient sparsification, in-network computing",

author = "Nasrin Razmi and Sourav Mukherjee and Bho Matthiesen and Armin Dekorsy and Petar Popovski",

year = "2025",

month = mar,

day = "13",

doi = "10.1109/IEEECONF62907.2025.10949091",

language = "English",

isbn = "979-8-3315-2290-2",

pages = "1--6",

booktitle = "2025 14th International ITG Conference on Systems, Communications and Coding (SCC)",

publisher = "IEEE (Institute of Electrical and Electronics Engineers)",

address = "United States",

note = "2025 14th International ITG Conference on Systems, Communications and Coding (SCC) ; Conference date: 10-03-2025 Through 13-03-2025",

}

Razmi, N, Mukherjee, S, Matthiesen, B, Dekorsy, A & Popovski, P 2025, Sparse Incremental Aggregation in Satellite Federated Learning. in 2025 14th International ITG Conference on Systems, Communications and Coding (SCC)., 10949091, IEEE (Institute of Electrical and Electronics Engineers), pp. 1-6, 2025 14th International ITG Conference on Systems, Communications and Coding (SCC), 10/03/2025. https://doi.org/10.1109/IEEECONF62907.2025.10949091

Sparse Incremental Aggregation in Satellite Federated Learning. / Razmi, Nasrin; Mukherjee, Sourav; Matthiesen, Bho et al.
2025 14th International ITG Conference on Systems, Communications and Coding (SCC). IEEE (Institute of Electrical and Electronics Engineers), 2025. p. 1-6 10949091.

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

TY - GEN

T1 - Sparse Incremental Aggregation in Satellite Federated Learning

AU - Razmi, Nasrin

AU - Mukherjee, Sourav

AU - Matthiesen, Bho

AU - Dekorsy, Armin

AU - Popovski, Petar

PY - 2025/3/13

Y1 - 2025/3/13

N2 - This paper studies Federated Learning (FL) in low Earth orbit (LEO) satellite constellations, where satellites are connected via intra-orbit inter-satellite links (ISLs) to their neighboring satellites. During the FL training process, satellites in each orbit forward gradients from nearby satellites, which are eventually transferred to the parameter server (PS). To enhance the efficiency of the FL training process, satellites apply in-network aggregation, referred to as incremental aggregation. In this work, the gradient sparsification methods from [1] are applied to satellite scenarios to improve bandwidth efficiency during incremental aggregation. The numerical results highlight an increase of over 4 x in bandwidth efficiency as the number of satellite in the orbital plane increases.

AB - This paper studies Federated Learning (FL) in low Earth orbit (LEO) satellite constellations, where satellites are connected via intra-orbit inter-satellite links (ISLs) to their neighboring satellites. During the FL training process, satellites in each orbit forward gradients from nearby satellites, which are eventually transferred to the parameter server (PS). To enhance the efficiency of the FL training process, satellites apply in-network aggregation, referred to as incremental aggregation. In this work, the gradient sparsification methods from [1] are applied to satellite scenarios to improve bandwidth efficiency during incremental aggregation. The numerical results highlight an increase of over 4 x in bandwidth efficiency as the number of satellite in the orbital plane increases.

KW - Federated learning

KW - Load modeling

KW - Low earth orbit satellites

KW - Numerical models

KW - Orbits

KW - Satellite constellations

KW - Satellites

KW - Servers

KW - Spectral efficiency

KW - Training

KW - Satellite Constellation

KW - gradient sparsification

KW - in-network computing

UR - http://www.scopus.com/inward/record.url?scp=105003533815&partnerID=8YFLogxK

U2 - 10.1109/IEEECONF62907.2025.10949091

DO - 10.1109/IEEECONF62907.2025.10949091

M3 - Article in proceeding

SN - 979-8-3315-2290-2

SP - 1

EP - 6

BT - 2025 14th International ITG Conference on Systems, Communications and Coding (SCC)

PB - IEEE (Institute of Electrical and Electronics Engineers)

T2 - 2025 14th International ITG Conference on Systems, Communications and Coding (SCC)

Y2 - 10 March 2025 through 13 March 2025

ER -

Sparse Incremental Aggregation in Satellite Federated Learning

Abstract

Conference

Keywords

Access to Document

AUB Link

Other files and links

Fingerprint

Sparse Incremental Aggregation in Satellite Federated Learning

Cite this