Hierarchical RBF-KAN and RBF-SKAN Architectures for Multidimensional Function Approximation and Random Field Learning
About
In this manuscript, we propose and analyze hierarchical Kolmogorov--Arnold neural network architectures employing radial basis functions as activation functions for approximating deterministic functions and random field models. Specifically, we develop a hierarchical radial-basis-function Kolmogorov--Arnold network (hierarchical RBF-KAN) for multidimensional deterministic function approximation and a hierarchical radial-basis-function stochastic Kolmogorov--Arnold network (hierarchical RBF-SKAN) for random field learning. From a theoretical perspective, we establish universal approximation results for both architectures. In particular, we derive quantitative approximation estimates for the hierarchical RBF-KAN, showing that the proposed framework has the potential to partially alleviate the curse of dimensionality in learning high-dimensional functions by reducing the effective dimensionality of the approximation problem. Furthermore, we show that the hierarchical RBF-SKAN can approximate random field models under the Wasserstein-2 metric. Empirically, we show that our proposed radial-basis-function-based neural network structure could effectively learn multivariate functions and random field models.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Random Field Reconstruction | Random Field Model Example 3 Eq. (29) (test) | Avg Relative Error (mean)0.0449 | 30 | |
| Function Approximation | Dimension 1 | Runtime15 | 27 | |
| Function Approximation | Oscillatory Function Equation 23 (test) | Error (d=1)0.0076 | 14 | |
| Function Approximation | Eq. (23) Dimension 2 | Runtime40.3 | 6 | |
| Random Field Reconstruction | Random field model d=1 Eq. 29 (test) | Training Runtime (s)541 | 6 | |
| Random Field Reconstruction | Random field model d=2 Eq. 29 (test) | Training Runtime (s)291 | 6 | |
| Random Field Reconstruction | Random field model d=3 Eq. 29 (test) | Training Runtime (s)730 | 6 | |
| Random Field Reconstruction | Random field model d=4 Eq. 29 (test) | Training Runtime (s)2.18e+3 | 6 | |
| Random Field Reconstruction | Random field model d=5 Eq. 29 (test) | Training runtime (s)4.24e+3 | 6 | |
| Function Approximation | Eq. (23) Dimension 3 | Runtime44.8 | 4 |