KAN/H: Kolmogorov-Arnold Network using Haar-like bases
This is an incremental improvement for researchers and practitioners in neural networks, potentially simplifying model deployment by reducing hyper-parameter tuning efforts.
The paper tackled the problem of hyper-parameter tuning in Kolmogorov-Arnold Networks by proposing KAN/H, which uses a Haar-like basis system instead of B-spline, and demonstrated its application to function approximation and MNIST without requiring most problem-specific tunings.
This paper proposes KAN/H, a variant of Kolmogorov-Arnold Network (KAN) that uses a Haar-variant basis system having both global and local bases instead of B-spline. The resulting algorithm is applied to function approximation problems and MNIST. We show that it does not require most of the problem-specific hyper-parameter tunings.