Strong proximinality and polyhedral spaces (Q5949004)

A closed subspace \(Y\) of a Banach space \(X\) is proximinal if every point of \(X\) has a nearest point in \(Y\). For such subspaces it does not necessarily follow that ``almost'' nearest points behave continuously in the following sense: given \(x \in X\) and \(\delta >0\) there is \(\varepsilon >0\) such that for all \(y_{\varepsilon}\) with \(\|x - y_{\varepsilon}\|< d(x,Y) + \varepsilon\) we have that \(y_{\varepsilon}\) is within \(\delta\) of the set of nearest points to \(x\). Spaces with this extra property are said to be strongly proximinal. This paper studies those Banach spaces for which every finite codimensional proximinal subspace is strongly proximinal. The main theorem uses the idea of a quasipolyhedral (QP) point. This notion was introduced by \textit{D. Amir} and \textit{L. Deutsch} [J. Approximation Theory 6, 176-201 (1972; Zbl 0238.41014)] and investigated extensively by \textit{R. Durier} and \textit{P. L. Papini} [Rocky Mt. J. Math. 23, No. 3, 863-875 (1993; Zbl 0799.46015)]. Main theorem: Let \(X\) be a Banach space with the property that every linear functional of norm \(1\) that attains its norm is a QP point of \(X^*\). If \(Y\) is a closed, finite codimensional subspace of \(X\) then the following are equivalent: (i) \(Y\) is strongly proximinal in \(X\); (ii) every linear functional that annihilates \(Y\) attains its norm. A metric projection is a choice of nearest point for each \(x \in X\). In the last section the authors show that for a strongly proximinal subspace one can make such a selection of first Baire class. If, instead, \(Y\) is a closed, proximinal subspace of finite codimension and such that the annihilator of \(Y\) is polyhedral, then one has a continuous selection.