Magic efficiency of the approximation of smooth functions by weighted means of two \(N\)-point Padé approximants (Q2319567)

The authors study extensively numerical examples for the approximation of a function knowing \(p_1>1,p_2,\ldots,p_N\) coefficients of its Taylor series at the points \(x_1,x_2,\ldots,x_n\). They assume the function \(f\) to have the`two-sided estimate property' (TSE): computing neighbouring \(N\)-point approximants \(f_1=[m/n]\) and \(f_2=[m-1n]\), these two approximants (the second calculated by removing one coefficient from the Taylor expansion of \(f\)) bound \(f\) in each interval \([x_i,x_{i+1}]\) from below resp. above. For a more detailed description of the procedure, the reader is referred to \S1 and to the paper where the method was presented originally: [\textit{R. Jedynak} and \textit{J. Gilewicz}, Ukr. Math. J. 65, No. 10, 1566--1576 (2014; Zbl 1309.41008)]. The layout of the paper is as follows: \S1. Introduction (\(1\) page) \S2. Inequalities between a Stieltjes function and its \(N\)-point Padé approximants (\(3\) pages) -- the Stieltjes function \(\frac{\ln{(1+x)}}{x}\) is used as an example. \S3. Convexity of the weight function (\(2\) pages) -- the same example as in the previous section is used. \S4. The TSE property, rescaling and modification of the approximated function (\(1\) page) \S5. Examples (\(11\frac{1}{2}\) pages) -- \(\mathcal{L}^{-1}(x)\), where \(\mathcal{L}(x)=\coth{x}-1/x\) (the Langevin function), \(R_1(x)=\mathcal{L}^{-1}(x)/x\), \(R_2(x)=(1-x)\mathcal{L}^{-1}(x)\) , \(R_3(x)=(1-x)\mathcal{L}^{-1}(x)/x\) and the Gaussian distribution. \S6. Conclusion (\(1/2\) page) References (\(21\) items)