Improving Generalizability of Kolmogorov-Arnold Networks via Error-Correcting Output Codes

TL;DR

This paper introduces a novel integration of Error-Correcting Output Codes into Kolmogorov-Arnold Networks, significantly improving their robustness and accuracy in multi-class medical image classification tasks.

Contribution

It is the first to combine ECOC with KAN, enhancing multi-class classification performance and generalizability in healthcare AI applications.

Findings

ECOC integration improves KAN accuracy on blood cell classification.

ECOC enhances robustness across different KAN variants.

The approach outperforms vanilla KAN in diverse settings.

Abstract

Kolmogorov-Arnold Networks (KAN) offer universal function approximation using univariate spline compositions without nonlinear activations. In this work, we integrate Error-Correcting Output Codes (ECOC) into the KAN framework to transform multi-class classification into multiple binary tasks, improving robustness via Hamming distance decoding. Our proposed KAN with ECOC framework outperforms vanilla KAN on a challenging blood cell classification dataset, achieving higher accuracy across diverse hyperparameter settings. Ablation studies further confirm that ECOC consistently enhances performance across FastKAN and FasterKAN variants. These results demonstrate that ECOC integration significantly boosts KAN generalizability in critical healthcare AI applications. To the best of our knowledge, this is the first work of ECOC with KAN for enhancing multi-class medical image classification performance.