Dependence on p of the best \(L^ p\) approximation operator (Q1089546)

Let (X,\({\mathcal B},\mu)\) be a measure space, V a finite dimensional subspace of \(L^ 1(X,{\mathcal B},\mu)\cap L^{\infty}(X,{\mathcal B},\mu),\) and \(f\in \{L^ 1(X,{\mathcal B},\mu)\cap L^{\infty}(X,{\mathcal B},\mu)\}\setminus V.\) For \(1<p<\infty\) define the best \(L^ p\) approximation operator at p, \(\tau_ f(p)\), to be the unique best approximation to f from V in the \(L^ p\)-norm. That is, \(\tau_ f(p)=v_ p\) with \(\| f-v_ p\|_ p=\min \{\| f-v\|_ p: v\in V\},\) where \(\| h\|_ p=\{\int_{X}| h|^ pd\mu \}^{1/p}.\) For \(p=1\) and \(p=\infty\) define \(\tau_ f(p)\) to be the set of best approximations to f in the corresponding \(L^ p\)-norm from V. Finally, set \(N_ f(p)=\| f-\tau_ f(p)\|_ p\) for \(1<p<\infty\) and \(N_ f(p)=\| f-h\|_ p,\) \(h\in \tau_ f(p)\) for \(p=1\) or \(p=\infty\). The authors study the best approximation operator \(\tau_ f(p)\) considering f fixed and p variable and illustrate their results by several examples. Concerning \(N_ f(p)\), they conjecture that it is a concave function of \(p\in (1,\infty)\).