On modified Mellin-Gauss-Weierstrass convolution operators (Q2136149)

The authors define two different modified Mellin-Gauss-Weierstrass convolution operators, that both preserve affine and quadratic functions. Let \(a_{w}(s)=se^{-2/w^2}\) and \(b_{w}(s)=se^{-1/w^2}\) be two parametric functions, such operators are of the following form \[ (\mathcal{M}_{w}^{a}f)(s):=\frac{w}{2\sqrt{\pi}}\int_{0}^{\infty}e^{-\left(\frac{w}{2}\ln t\right)^2}f(t a_{w}(s))\frac{dt}{t}, \] \[ (\mathcal{M}_{w}^{b}f)(s):=s b_{w}^{-1}(s)\frac{w}{2\sqrt{\pi}}\int_{0}^{\infty}e^{-\left(\frac{w}{2}\ln t\right)^2}f(t b_{w}(s))\frac{dt}{t^2}, \] with \(s\in\mathbb{R}^{+}\), \(w>0\), and where \(f\in L_p(\mu)\), \(1<p\leq +\infty\) i.e., \(f\) belongs to the Lebesgue spaces of \(p\)-integrable functions with respect to the Haar measure \(\mu\). The authors prove that, in terms of pointwise approximation error and convergence rate, these operators outperform the Mellin-Gauss-Weierstrass convolution operators [\textit{P. L. Butzer} et al., J. Fourier Anal. Appl. 3, No. 4 325--376 (1997; Zbl 0885.44004)]. Also, weighted approximation and quantitative Voronovskaja type formulas are investigated throughout the paper. Finally, they give some numerical results using the Mathematica software, which allows them to interpret the aforesaid results.