Multivariate approximation by a combination of modified Taylor polynomials (Q2499803): Difference between revisions

Let \(\Omega\) be a compact convex subset of \(\mathbb{R}^d\) \((d\geq 1)\), \(x_1,\dots, x_n\), \(n\) distinct points of \(\Omega\), and \(f\) a real valued function on \(\Omega\), \(m+ r+ 1\) times continuously differentiable \((m,r\in\mathbb{N})\). Let \(g_1,\dots,g_n\) real valued continuous functions on \(\Omega\) such that for every polynomial \(P\) of degree at most \(m\) we have \(P(x)= \sum^n_{j=1} P(x_j)g_j(x)\). Suppose that the values of \(f\) and those of its partial derivatives up to order \(r\) at \(x_1,\dots, x_n\) are known. This paper deals with various aspects of an approximation of \(f\) by an operator of the form \(\sum^n_{j=1} T_r[f, x_j](x) g_j(x)\), where \(T_r[f, x_j]\) is the Taylor polynomial of \(f\) in \(x_j\) of degree \(r\) with the coefficients appropriately modified. In particular, the case when \(g_1,\dots, g_n\) is a partition of unity and the case of multivariate Bernstein polynomials are widely studied.