KANs achieve higher or comparable accuracy to MLPs across four tabular datasets in simulated federated learning, using fewer communication rounds.
Evaluating Federated Kolmogorov-Arnold Networks on Non-IID Data
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
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.
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Enhancing Federated Learning with Kolmogorov-Arnold Networks: A Comparative Study Across Diverse Aggregation Strategies
KANs achieve higher or comparable accuracy to MLPs across four tabular datasets in simulated federated learning, using fewer communication rounds.