Search papers, labs, and topics across Lattice.
This study empirically benchmarks Kolmogorov-Arnold Networks (KANs) against Multi-Layer Perceptrons (MLPs) for structured data classification across twelve datasets, evaluating their performance on binary, multiclass, multilabel, and ordinal tasks. The results indicate that KANs significantly outperform MLPs in terms of accuracy and F1-Score, although the medium effect size suggests a trade-off between performance and increased computational complexity. Consequently, while KANs are advantageous for high-precision applications, MLPs remain a viable choice for scenarios with limited resources.
KANs deliver superior classification performance but come with a hefty computational price tag that may not justify their use in all scenarios.
This study presents an empirical benchmarking comparison between Kolmogorov-Arnold Networks (KANs) and Multi-Layer Perceptrons (MLPs) on structured tabular classification tasks. Motivated by the growing interest in KANs as an alternative function-approximating architecture, we evaluate their out-of-the-box performance on twelve publicly available datasets spanning binary, multiclass, multilabel, and ordinal problems. Both models were trained under standardized preprocessing, architecture, and fixed hyperparameter settings, with performance assessed using test accuracy and F1-Score, paired hypothesis testing, and effect size analysis. Results show that KANs statistically outperform MLPs in binary and multiclass domains and achieve a significant aggregate advantage across all datasets. However, the observed medium effect size (d = -0.46) raises an important cost-benefit consideration: while KANs offer superior generalization through adaptive spline-based mappings, this advantage comes with substantially higher parameter and computational complexity relative to the MLP baseline. These findings suggest KANs are the preferred choice for high-precision applications, while MLPs remain a robust and efficient option for resource-constrained environments. Future work should extend this analysis to additional data modalities to further refine these architectural selection criteria.
