Optimal Approximation by $sk$-Splines on the Torus

Abstract

Fixed a continuous kernel K on the $d$-dimensional torus, we consider a generalization of the univariate $sk$-spline to the torus, associated with the kernel K. It is proved an estimate which provides the rate of convergence of a given function by its interpolating $sk$-splines, in the norm of $L^q$ for functions of the type $f=K\ast \phi$ where $\phi \in L^p$ and $1\le p\le 2\le q\le \infty ,1/p-1/q\ge 1/2$. The rate of convergence is obtained for functions f in Sobolev classes and this rate gives optimal error estimate of the same order as best trigonometric approximation, in a special case.

28 pages