Analytic monotone pseudospectral interpolation (Q895424): Difference between revisions

Given an interpolating table \( U = (x _j , y_ j )_{j\in M}\); \(M=\{ 1,2, \dots, m\}\), where \(\{x_j \}_{j\in M} \) and \(\{y_j \}_{j\in M}\) are strictly monotonic increasing sequences with \(x_1 = 0\); \(x_m = 2\pi\), the object of the paper is to construct a smooth approximant that preserves the monotonicity of the data. Thus, the author constructs an analytic, monotone, and periodic function \(f\) such that \(|f(x_ j) - y_ j|\), \(j \in M\), is bounded by some function \(\epsilon (m)\), which tends to zero as \(m \rightarrow \infty\). The resulting approximants are either algebraic polynomials, or linear functions, \(L\), plus trigonometric polynomials. In both cases the least possible degree is either an absolute constant, or \(\ln m\), multiple of the least degree in [\textit{G. L. Iliev}, J. Approx. Theory 28, 101--112 (1980; Zbl 0425.41006)].