Sparse Incremental Aggregation in Satellite Federated Learning

Nasrin Razmi 1 2 , Sourav Mukherjee 1 2 , Bho Matthiesen 1 2 , Armin Dekorsy 1 2 , Petar Popovski 3 1 · 2025

Read Sparse Incremental Aggregation in Satellite Federated Learning by Nasrin Razmi 1 2 , Sourav Mukherjee 1 2 , Bho Matthiesen 1 2 , Armin Dekorsy 1 2 , Petar Popovski 3 1 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-11385.txt.

Description

Razmi and co-authors study in 2025 how to train shared models across satellite fleets without drowning the links between craft. In low Earth orbit, each satellite passes its learning updates to neighbors along the ring, and those updates pile up on the way to the ground server. Trimming updates in transit, a trick called incremental aggregation, already cuts the load.

Here the team adds sparsification, sending only the biggest update entries, on top of that pipeline. Their trials show bandwidth gains of more than fourfold as the number of satellites in a ring grows. The work speaks to learning swarms that must share thin space links.

Catalog record

CREATORNasrin Razmi 1 2 , Sourav Mukherjee 1 2 , Bho Matthiesen 1 2 , Armin Dekorsy 1 2 , Petar Popovski 3 1
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE27 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-11385

Browse the full library · About the Tower of Babel