Uniform convergence of Nyström discretization on Hölder spaces

From MaRDI portal
Publication:2168959





The discretization of nonlocal operators has attracted the interest of many researchers since this allows the simulation and the practical application of models with nonlocal nature, which have become increasingly popular in physics, engineering and other fields of science. The author studies Nyström-type discretizations of linear and nonlocal Fredholm-type operators \[ {\mathcal K} u := \int_\Omega k(\cdot,y) u(y)\, d y, \tag{1} \] in the context of Hölder regularity. Here the integration kernel is supposed to be known. The context of Hölder spaces seems to be suitable as far as this framework typically allows the study of different kind of problems under optimal regularity requirements. In the present work the Hölder regularity is understood for the integration kernel \(k\). The author considers a naive generalization of the classical quadrature rules (rectangle, midpoint, trapezoidal) of the so-called Nyström methods for weighted integrals of type (1): \[ {\mathcal K}^n u := \sum_{\nu\in\Omega^{n}} w_\nu k(\cdot, \nu) u(\nu), \] where \(\Omega^n\) is a grid on \(\Omega\) and \(w_\nu\) stand for the integration weights in the counterpart classical quadrature. Afterwards the convergence of such a kind of quadratures is stated in Theorems 4 and 7 under suitable Hölder regularity assumptions on \(k\). The error bounds show a convergence rate (order), and in addition show a precise description of error constants, not only in terms of the regularity assumptions on \(k\) but also in terms of the domain \(\Omega\). Finally the author applies the theoretical achievements for general equations to a naive convolution equation. At this point some numerical experiments might be expected by the readers.











This page was built for publication: Uniform convergence of Nyström discretization on Hölder spaces

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2168959)