Uniqueness in vector-valued approximation (Q1801570)

By vector-valued functions one means the mappings \(f(x):= (f_ 1(x),\dots, f_ m(x))\), \(x\in D\), where \(D\) is some set and each \(f_ i\) \((i=1,\dots,m)\) is defined on \(D\) with real values. One can consider the mixed norms of such functions given by \[ \| f\|_{A(p,q)}:= \left( \sum^ m_{i=1} \biggl( \int_ D | f_ i(x)|^ q d\nu(x)\biggr)^{p/q} \right)^{1/p} \] and \[ \| f\|_{B(p,q)}:= \left(\int_ D \left( \sum^ m_{i=1} | f_ i(x)|^ q\right) ^{p/q} d\nu(x)\right) ^{1/p} \] where \(p,q\in[1,\infty]\). The author gives in the main an answer at the following question: Given a finite dimensional subspace \(U\), what are conditions on \(U\) such that to each \(f\) there exists a unique best approximant from \(U\) in the norm \(\|\cdot\|_{A(p,q)}\) or \(\|\cdot\|_{B(p,q)}\)? The paper contains three sections. After an introductory paragraph, he gives a series of known results concerning characterizing best approximations and uniqueness. These results are used in Section 3 which presents a review of some results on unicity spaces in the \(C\) and \(L^ 1\) norms, as these are the basic non-smooth, non-strictly convex norms considered. Sections 4-12 represent Part A and are concerned with \(A(p,q)\)-norms and Sections 13-19 represent Part B and deal with the \(B(p,q)\)-norms.