Approximation by extreme functions (Q1288260): Difference between revisions

For a topological space \(T\) and a real normed space \(X\) let \(Y= C(T,X)\), the normed space of all \(X\)-valued bounded continuous functions on \(T\) endowed with the supremum norm. Let \(Y^{-1}=\{f\in Y:f\) does not vanish in any \(t\in T\}\). For \(f\in Y\), let \(\alpha(f)= d(f,Y^{-1})\) and let \(m(f)= \inf \{\|f(t)\|: t\in T\}\). First, it is shown that \(\alpha(f)\) is the infimum of those \(\delta>0\) for which there exists a continuous extension \(e\) from \(T\) to the unit sphere \(S(X)\) of \(X\) of the continuous function \(t\to{f(t) \over \|f(t)\|}\), \(\|t\|\geq\delta\). This is used to prove that any \(f\in Y\) has a best approximation in \(Y^{-1}\) which can be chosen to have the least possible norm. Now let \(X\) be further strictly convex. Let \(E(Y)\) be the set of extreme points of the unit ball of \(Y\). Then it is shown that \(d(f,E (Y)) =\max\{1-m(f) +\alpha(f)\), \(\|f\|-1\}\), \(f\in Y\). Then it is deduced that \(E(Y)\) is proximinal in \(Y^{-1}\) and sufficient conditions are given under which \(f\in Y\setminus Y^{-1}\) admits a best approximation. The last section is devoted to the study of the Aron-Lohman \(\lambda\)-function of \(Y\).