H-sets for kernel-based spaces

DOI10.1016/J.JAT.2023.105942MaRDI QIDQ6093306zbMATH OpenOpenAlexFDO

Publication date 6 September 2023

Published in Journal of Approximation Theory (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2107.10040

alternation approximation duality error bounds kernels radial basis functions reproducing kernel Hilbert spaces stability uniqueness

Mathematics Subject Classification ID

Approximation by polynomials (41A10) Uniqueness of best approximation (41A52) Algorithms for approximation of functions (65D15)

Abstract: The concept of

H

-sets as introduced by Collatz in 1956 was very useful in univariate Chebyshev approximation by polynomials or Chebyshev spaces. In the multivariate setting, the situation is much worse, because there is no alternation, and

H

-sets exist, but are only rarely accessible by mathematical arguments. However, in Reproducing Kernel Hilbert spaces,

H

-sets are shown here to have a rather simple and complete characterization. As a byproduct, the strong connection of

H

-sets to Linear Programming is studied. But on the downside, it is explained why

H

-sets have a very limited range of applicability in the times of large-scale computing.

Review text

H-sets are a classical instrument going back to Collatz to generalise the approach to characterising and computing (by the so-called exchange algorithm) univariate best uniform approximants (Chebyshev approximations). The purpose of this article is to generalise this concept further, namely to more than one variables. Essentially, the H-set conditions pose a requirement on the point set \(H\subseteq T\) that enables the classical estimates in the characterisation of best uniform approximations to hold. That is, that there is no element \(v\) from the set of approximants \(V\subseteq C(T)\) that has the property that \(\operatorname{sign}(h)v(h)<0\) for all \(h\in H\). This implies for an approximation \(\tilde v\in V\) \[ \|f-v\|_\infty \geq (f(h)-v(h))\operatorname{sign}(h) = (f(h)-\tilde v(h))\operatorname{sign}(h)+(\tilde v(h)-v(h))\operatorname{sign}(h) \] must be equal to \[ (f(h)-\tilde v(h))\operatorname{sign}(h), \] and so \[ \inf_{v\in V}\|f-v\|_\infty \geq (f(\tilde h)-\tilde v(\tilde h))\operatorname{sign}(\tilde h)\geq\inf_{h\in H}(f(h)-\tilde v(h))\operatorname{sign}(h) \] for some \(\tilde h \in H\). In this paper, first, an equivalent formulation of the H-set property in terms of a linear optimisation problem with simple constraints is established. This can then be used for the numerical computation when we wish to verify the H-set property. Also, second, a multivariable equivalent of the H-set property for kernel-based (e.g., radial basis functions) approximation spaces is given, also as a consequence of the mentioned linear optimisation problem, including a numerical example.

Recommendations

Cites work

Cited in

(2)

This page was built for publication: \(H\)-sets for kernel-based spaces

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6093306)