Kolmogorov-Arnold Networks are Radial Basis Function Networks
About
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.
Ziyao Li• 2024
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Human Activity Recognition | PAMAP2 | -- | 54 | |
| Human Activity Recognition | Opportunity | Macro F143.4 | 43 | |
| Activity Recognition | mHealth | -- | 35 | |
| Human Activity Recognition | SKODA | Macro F188.7 | 29 | |
| Human Activity Recognition | MotionSense | Macro-F182.8 | 29 | |
| Human Activity Recognition | HAPT | Macro-F167.3 | 20 | |
| Activity Recognition | DSADS | Macro F148.6 | 20 | |
| Activity Recognition | DG | Macro F1 Score56.2 | 20 | |
| Physics-Informed PDE Solving | 2D Helmholtz (test) | Time (sec/iter)0.0228 | 14 | |
| Function Approximation | Function Approximation [np, 6, 1] | Time (sec/iter)0.0305 | 14 |
Showing 10 of 14 rows