Evaluating Federated Kolmogorov-Arnold Networks on Non-IID Data

Arthur Mendonça Sasse; Claudio Miceli de Farias

Evaluating Federated Kolmogorov-Arnold Networks on Non-IID Data

Machine Learning 2024-10-14 v1 Artificial Intelligence

Authors: Arthur Mendonça Sasse , Claudio Miceli de Farias

Abstract

Federated Kolmogorov-Arnold Networks (F-KANs) have already been proposed, but their assessment is at an initial stage. We present a comparison between KANs (using B-splines and Radial Basis Functions as activation functions) and Multi- Layer Perceptrons (MLPs) with a similar number of parameters for 100 rounds of federated learning in the MNIST classification task using non-IID partitions with 100 clients. After 15 trials for each model, we show that the best accuracies achieved by MLPs can be achieved by Spline-KANs in half of the time (in rounds), with just a moderate increase in computing time.

Keywords

Federated Kolmogorov-Arnold Networks non-IID data MNIST classification B-splines Radial Basis Functions

Cite

@article{arxiv.2410.08961,
  title  = {Evaluating Federated Kolmogorov-Arnold Networks on Non-IID Data},
  author = {Arthur Mendonça Sasse and Claudio Miceli de Farias},
  journal= {arXiv preprint arXiv:2410.08961},
  year   = {2024}
}

Comments

10 pages, 5 figures, for associated code see https://github.com/artsasse/fedkan

Related papers

View all related →