Uniqueness of best simultaneous approximation and strictly interpolating subspaces (Q789632): Difference between revisions

If F,G are subsets of a normed linear space E,F bounded, a best simultaneous approximation to F in G is \(y_ 0\in G\) minimizing \(r(y,G)=\sup_{x\in F}\| x-y\|\). The set of best simultaneous approximations to F in G is denoted \(Z_ G(F)\) and called also the ''Chebyshev center'' for F in G, and \(r_ G(F)=\inf_{y\in G}r(y,F)\) is called the ''Chebyshev radius'' of F with respect to G. A weak uniqueness property for Chebyshev centers had been observed by Golomb and others for interpolating subspaces. If \(y_ 1,...,y_ n\) are linearly independent elements of a normed space E, call \(G=\{\sum^{n}_{i=1}c_ iy_ i;c_ i\in J_ i\}\) an ''RS-set'' if \(J_ i\) are intervals of the types (I) a single point, (II) a nontrivial proper closed (bounded or unbounded) interval in R, or (III) the whole line, and if every subset of \(\{y_ 1,...,y_ n\}\) consisting of all \(y_ i\) with \(J_ i\) of type III and some \(y_ i\) with \(J_ i\) of type II spans an interpolating subspace (in particular, any interpolating subspace is an RS-set). Consider the Theorem: If G is an RS-set in a normed linear space E, then for every compact \(F\subset E\) with \(r_ E(F)<r_ G(F)\), \(Z_ G(F)\) is a single point. It was proved by Rozema and Smith for the case \(E=C[a,b]\). They suggested its generalization to the more general setting. \textit{J. M. Lambert, D. D. Milman} [J. Approximation Theory 26, 71-78 (1979; Zbl 0401.41037)] claimed that it is valid even when F is assumed only to be bounded (and not necessarily compact). In the present paper it is shown that this is true only in the compact case. A corrected version for the noncompact case is obtained when one replaces, in the definition of an ''interpolating subspace'', the set of extreme points of the dual ball by its weak closure, getting thus a ''strictly interpolating subspace'' and a ''strictly RS-set''.