Canonical sets of best \(L_1\)-approximation (Q1757910)

Let \(K\subset\mathbb{R}^{d}\) be a compact set, let \(C(K)\) be the linear space of continuous functions on \(K,\) and let \(\mu\) be a positive Borel measure defined on \(K\) and \(\left\| f\right\| _{1}:=\int_{K}\left| f\right| d\mu\) the \(L_{1}(\mu)\)-norm on \(C(K).\) Let \(\mathcal{U}\) be a subspace of \(C(K)\). A subset \(X\subset K\) is called unisolvent for \(\mathcal{U}\) if the interpolation problem \((\ast)\;u(x)=f(x)\), \(x\in X\), possesses a unique solution \(u\in\mathcal{U}\) for every \(f\in C(K)\). If \(\mathcal{C}\subset C(K)\) is a class of functions, a set \(X\subset K\) which is unisolvent for \(\mathcal{U}\) is a canonical set of best \(L_{1}(\mu )\)-approximation to \(\mathcal{C}\) if, for all \(f\in\mathcal{C}\), the solution of interpolation problem \((\ast)\) is a best \(L_{1}(\mu)\)-approximant to \(f\in\mathcal{C}\) from \(\mathcal{U}\). In the case of best one-sided approximation, the interpolation problem is considered to be in the sense of Lagrange-Hermite interpolation, i.e., using not only the values of \(u\) and \(f\) but also the values of their first (partial) derivatives. In this paper, the authors present results on canonical sets of best \(L_{1}(\mu)\)-approximation for univariate and multivariate interpolation and best \(L_{1}(\mu)\)-approximation by blending functions.