Showing 1–1 of 1 results for author: Audone, G

A Recipe for Learning Variably Scaled Kernels via Discontinuous Neural Networks

Authors: Gianluca Audone, Francesco Della Santa, Emma Perracchione, Sandra Pieraccini

Abstract: The efficacy of interpolating via Variably Scaled Kernels (VSKs) is known to be dependent on the definition of a proper scaling function, but no numerical recipes to construct it are available. Previous works suggest that such a function should mimic the target one, but no theoretical evidence is provided. This paper fills both the gaps: it proves that a scaling function reflecting the target one… ▽ More The efficacy of interpolating via Variably Scaled Kernels (VSKs) is known to be dependent on the definition of a proper scaling function, but no numerical recipes to construct it are available. Previous works suggest that such a function should mimic the target one, but no theoretical evidence is provided. This paper fills both the gaps: it proves that a scaling function reflecting the target one may lead to enhanced approximation accuracy, and it provides a user-independent tool for learning the scaling function by means of Discontinuous Neural Networks ($δ$NN), i.e., NNs able to deal with possible discontinuities. Numerical evidence supports our claims, as it shows that the key features of the target function can be clearly recovered in the learned scaling function. △ Less

Submitted 20 May, 2025; v1 submitted 15 July, 2024; originally announced July 2024.

MSC Class: 65D15; 1A05; 68Q32

Journal ref: Journal of Computational and Applied Mathematics, Volume 469, 2025, 116653 Journal of Computational and Applied Mathematics, Volume 469, 116653