Fed-KAN: Federated Learning With Kolmogorov–Arnold Networks for Traffic Prediction: KAN-Based Learning
IEEE Vehicular Technology Magazine · 2026-06 · 21 (2) · 40-48
DOI: 10.1109/mvt.2026.3679100 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Vehicular Technology Magazine, 2026-06. Verfasst von Engin Zeydan und 5 weiteren Autor:innen. Die Arbeit nennt 15 Literaturangaben. Gefördert von: European Union’s Horizon Europe Research and Innovation program.
Abstract (Originalsprache)
Nonterrestrial networks (NTNs) are becoming a critical component of modern communication infrastructures, especially with the advent of low Earth orbit (LEO) satellite systems. Traditional centralized learning approaches face major challenges in such networks due to high latency, intermittent connectivity, and limited bandwidth. Federated learning (FL) is a promising alternative as it enables decentralized training while maintaining data privacy. However, existing FL models, such as FL with multilayer perceptrons (Fed-MLP), can struggle with high computational complexity and poor adaptability to dynamic NTN environments. This article provides a detailed analysis for FL with Kolmogorov–Arnold networks (Fed-KAN), its implementation, and performance improvements over traditional FL models in NTN environments for traffic forecasting. The proposed Fed-KAN is a novel approach that utilizes the functional approximation capabilities of KANs in an FL framework. We evaluate Fed-KAN compared to Fed-MLP on a traffic dataset of a real satellite operator and show a significant reduction in training and test loss. Our results show that Fed-KAN can achieve a 49% reduction in average test loss compared to Fed-MLP, highlighting its improved performance and better generalization ability. At the end of the article, we also discuss some potential applications of Fed-KAN within open radio access network (O-RAN) and Fed-KAN usage for split functionalities in NTN architecture.
Förderung
- European Union’s Horizon Europe Research and Innovation program
Zitieren
APA
Zeydan, E., Vaca-Rubio, C. J., Botana, L. B., Pereira, R., López, M. C., & Dev, K. (2026). Fed-KAN: Federated Learning With Kolmogorov–Arnold Networks for Traffic Prediction: KAN-Based Learning. IEEE Vehicular Technology Magazine, 21(2), 40-48. https://doi.org/10.1109/mvt.2026.3679100
Vancouver
Zeydan E, Vaca-Rubio CJ, Botana LB, Pereira R, López MC, Dev K. Fed-KAN: Federated Learning With Kolmogorov–Arnold Networks for Traffic Prediction: KAN-Based Learning. IEEE Vehicular Technology Magazine. 2026;21(2):40-48. doi:10.1109/mvt.2026.3679100
BibTeX
@article{Zeydan2026,
author = {Zeydan, Engin and Vaca-Rubio, Cristian J. and Botana, Luis Blanco and Pereira, Roberto and López, Màrius Caus and Dev, Kapal},
title = {Fed-KAN: Federated Learning With Kolmogorov–Arnold Networks for Traffic Prediction: KAN-Based Learning},
journal = {IEEE Vehicular Technology Magazine},
year = {2026},
volume = {21},
number = {2},
pages = {40--48},
doi = {10.1109/mvt.2026.3679100}
}
EUVIMED