Kolmogorov-Arnold Networks are Radial Basis Function Networks

TL;DR

This paper demonstrates that Kolmogorov-Arnold Networks can be approximated by Gaussian RBFs, leading to a faster implementation called FastKAN that retains the original network's properties.

Contribution

It introduces FastKAN, a novel RBF network approximation of KANs, significantly improving computational efficiency.

Findings

FastKAN is much faster than traditional KAN implementations.

Kolmogorov-Arnold Networks can be effectively approximated by Gaussian RBFs.

The approximation preserves the core properties of KANs.

Abstract

This short paper is a fast proof-of-concept that the 3-order B-splines used in Kolmogorov-Arnold Networks (KANs) can be well approximated by Gaussian radial basis functions. Doing so leads to FastKAN, a much faster implementation of KAN which is also a radial basis function (RBF) network.