A Comparative Study of Federated Learning Aggregation Strategies under Homogeneous and Heterogeneous Data Distributions
Antonios Makris 1 , Christos Dousis 2 , Emmanouil Kritharakis 1 , Stavros Bouras 1 , Konstantinos Tserpes 1 · 2026
Read A Comparative Study of Federated Learning Aggregation Strategies under Homogeneous and Heterogeneous Data Distributions by Antonios Makris 1 , Christos Dousis 2 , Emmanouil Kritharakis 1 , Stavros Bouras 1 , Konstantinos Tserpes 1 (2026) 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-2605-11010.txt.
Description
Antonios Makris and four colleagues compare ways to combine model updates in federated learning, where training runs across scattered devices. The server's combining rule strongly shapes learning quality, toughness, and system behavior, and its best choice may depend on how alike the devices' data are.
Using image-classification test sets, the team measures how widely used combining rules respond to varying data mix, tracking central accuracy and loss plus combining, training, and communication time. Rules show distinct trade-offs that shift with data set and operating conditions, so no single rule wins everywhere. Federated practitioners get an experimental map for matching the combining choice to data conditions.
Catalog record
| CREATOR | Antonios Makris 1 , Christos Dousis 2 , Emmanouil Kritharakis 1 , Stavros Bouras 1 , Konstantinos Tserpes 1 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36457 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.11010