Best approximation by free knot splines (Q5937105)

Denote by \(S_{m,n,r}\left( \xi,\lambda\right) ,\) \(m\in{\mathbb{N}} , m\geq 2, r\in{\mathbb{N}}\) the vector space of all polynomial spline functions \(s:{\mathbb{R}}\rightarrow {\mathbb{R}}\) of order \(m\) with knots \(\xi=\left( \xi_{\nu}\right) _{\nu=1}^{r}\) satisfying \(\xi_{1}<\xi_{2}<...<\xi_{n}\), and having the respective multiplicities \(\lambda_{v}\in{\mathbb{N}},\nu=1,2,...,r\) with \(\sum_{\nu=1}^{r}\lambda_{\nu}=n.\) Let also \(M_{m,n,r}\left( \lambda\right) =\bigcup\limits_{\xi}S_{m,n,r}\left( \xi,\lambda\right) \) be the set of polynomial spline functions of order \(m\) with free knots and total multiplicities \(n.\) The Gauss transform with parameter \(t>0\) maps the set \(M_{m,n,r}\) into a new set, denoted by \(G_{m,n,r}\), which satisfy the local Haar condition and has good structural properties. This allows to obtain some necessary and sufficient criteria for the best approximation elements with respect to the Chebyshev norm of the functions from \(C\left[ a,b\right] \) by elements of \(G_{m,n,r}\). Letting \(t \downarrow 0\) one obtains the best approximation in the original space of spline functions. Several numerical examples are included.