Random points are optimal for the approximation of Sobolev functions

Authors David Krieg, Mathias Sonnleitner

Publication date 23 September 2020

Rate of convergence, degree of approximation (41A25) Multidimensional problems (41A63) Sampling theory, sample surveys (62D05) Algorithms for approximation of functions (65D15) Numerical integration (65D30)

Abstract: We show that independent and uniformly distributed sampling points are as good as optimal sampling points for the approximation of functions from the Sobolev space

W_{p}^{s} (O m e g a)

on bounded convex domains

O m e g a s u b s e t m a t h b b R^{d}

in the

L_{q}

-norm if

q < p

. More generally, we characterize the quality of arbitrary sampling points

P s u b s e t O m e g a

via the

L_{g} a m m a (O m e g a)

-norm of the distance function

m d i s t (c d o t, P)

, where

g a m m a = s (1 / q - 1 / p)^{- 1}

if

q < p

and

g a m m a = i n f t y

if

q g e p

. This improves upon previous characterizations based on the covering radius of

P

.

This page was built for publication: Random points are optimal for the approximation of Sobolev functions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6349729)