The inverse problem of best approximation theory in \(F\)-spaces (Q1594367)

Let \(\{L_n\}_{n\in\mathbb{N}}\) be a sequence of subspaces of a space \(X\), \(L_n\subset L_{n+1}\). For each \(x\in X\) this sequence generates a nonincreasing sequence of numbers \(a_n= E(x,L_n)\), convergent to zero where \(E(x, L_n)\) is the best approximation of \(x\) in \(L_n\). The inverse problem of the best approximation theory can be stated as follows: Let \(\{a_n\}_{n\in\mathbb{N}}\) be a given nonincreasing convergent to zero sequence of nonnegative numbers and \(\{L_n\}_{n\in\mathbb{N}}\) a given sequence of subspaces of a space \(X\); there exists an element \(x\in X\) such that \(a_n= E(x, L_n)\)?. (Of course, the space \(X\) is endowed with a structure which allows to set the problem of the best approximation.) The author treats this problem in \(F\)-spaces, i.e. in complete real linear spaces a metric invariant under shift. Additionally, it is assumed that the metric \(d\) of the \(F\)-space \(X\) is monotone, i.e. \(x\in X\setminus\{\theta\}\) and \(0< \alpha< \beta\) imply \(d(\theta, dx)\leq d(\theta,\beta x)\), where \(\theta\) is the zero of \(X\). The problem in question is examined also in locally pseudo-convex \(F\)-spaces and in \(F\)-spaces of mappings of a measure space into an \(F\)-space.