Continuous selections and maximal alternators for spline approximation (Q1069077)

Let [a,b] be a closed interval, let k be a positive integer and suppose \(\Delta =\{a=x_ 0<x_ 1<...<x_ k<x_{k+1}=b\}\) is a partition of [a,b]. Given an integer \(m\geq 2\) \(S_ m(\Delta)\) stands for the space of polynomial splines of order m with (simple) knots of \(\Delta\). For any \(f\in [a,b]\), the set of uniform best approximations of f by elements of \(S_ m(\Delta)\) is denoted by P(f). If S: C[a,b]\(\to C[a,b]\) with Sf\(\in P(f)\) for all \(f\in C[a,b]\) then S is called a continuous selection for the set-valued metric projection P. The authors prove Theorem 1. There is never a unique continuous selection for spline approximation. The next part of the paper deals with the notion of maximal alternators. A nonzero function \(g\in C[a,b]\) is said to alternate p times (on \(p+1\) points) provided there exist \(a\leq t_ 1<t_ 2<...<t_{p+1}\leq b\) such that either \((-1)^ ig(t_ i)=\| g\|\), \(i=1,2,...,p+1\) or \((- 1)^{i+1}g(t_ i)=\| g\|\), \(i=1,2,...,p+1\). A spline \(s\in S_ m(\Delta)\) for which f-s alternates a maximal number of times is called a maximal alternator for f. Theorem 2. The space \(S_ m(\Delta)\) has the property that every \(f\in C[a,b]\sim S_ m(\Delta)\) has a unique maximal alternator if and only if \(k\leq m.\) Let \(M_ f\) be defined by \(M_ f=f\) if \(f\in S_ m(\Delta)\), and \(M_ f=s_ f\) otherwise. If \(P^*(f)=\{Sf: S\) is a continuous selection for the metric projection \(P\}\) then the following theorem holds: Theorem 3. For every \(f\in C[a,b]\), \(M_ f\in P^*(f)\).