Optimal order Jackson type inequality for scaled Shepard approximation (Q1697439): Difference between revisions

Let \(X\) be a convex domain in \(\mathbb{R}^d\) and let \(x_j\in X\), \(j=1,\dots,n\). Denote by \(BC(X)\) the set of bounded and continuous functions on \(X\). Given \(\lambda\geq 1\), the Shepard approximation of a function \(f\in BC(X)\) can be written as \[ f(x)\approx \frac{1}{\sum_{j=1}^n|x - x_j|^{-\lambda}}\sum_{j=1}^n\frac{f(x_j)}{|x-x_j|^{\lambda}}. \] The authors study a variation of the Shepard approximation by introducing a dilation factor into the base function \(|x|^{-\lambda}\), which synchronizes with the Hausdorff distance between the data set \(\{x_1,\dots,x_n\}\) and the domain \(X\). An optimal order Jackson type error estimate for all \(f\in BC(X)\) on any given convex domain \(X\) is established. Moreover, an upper bound estimate due to \textit{F. J. Narcowich} and \textit{J. D. Ward} [J. Approx. Theory 64, No. 1, 69--94 (1991; Zbl 0724.41004)] for the numbers of well-separated points in thin annuli is improved.