Existence of best simultaneous approximations in \(L_p (S,\, \varSigma,\, X)\) (Q719366): Difference between revisions

A norm \(N\) on \(\mathbb R^m\) is called monotonic if \(|a_i|\leq |b_i|,\;i=1,\dots,m,\, \) implies \(N(a)\leq N(b),\) for every \(a=(a_i)_{i=1}^m\) and \((b=(b_i)_{i=1}^m\) in \(\mathbb R^m.\) If \(X\) is a normed space and \(G\) a subset of \(X\), then an \(N\)-simultaneous best approximation element to \((x_1,\dots,x_m)\in X^m\) is an element \(g_0\in G\) such that \(N\left((\|f_i-g_0\|)_{i=1}^m\right)\leq N\left((\|f_i-g\|)_{i=1}^m\right) \) for every \(g\in G.\) If such an element exists for every \((x_1,\dots,x_m)\in X^m\), then the set \(G\) is called \(N\)-simultaneously proximinal in \(X\). The authors study this problem in the case of the space \(L_p(S,\Sigma,X)\) of \(p\)-integrable vector-functions (essentially bounded if \(p=\infty\)) with values in a Banach space \(X\) and \(G= L_p(S,\Sigma_0,Y),\) where for a sub-\(\sigma\)-algebra \(\Sigma_0\) of the \(\sigma\)-algebra \(\Sigma\) and a subset \(Y\) of \(X,\, L_p(S,\Sigma_0,Y)\) denotes the set of all \(g\in L_p(S,\Sigma_0,X)\) such that \(g(s)\in Y,\) a.e. \(s\in S.\) The main result of the paper (Theorem 3.1) asserts that if \(Y\) is a locally weakly compact closed convex subset of a Banach space \(X\) such that both \(\overline{span}(Y)\) and \(\left(\overline{span}(Y)\right)^*\) have the Radon-Nikodym property, then \(L_p(S,\Sigma_0,Y)\) is \(N\)-simultaneously proximinal in \(L_p(S,\Sigma,X),\, 1\leq p\leq \infty.\) Although, for the sake of simplicity, the authors restrict themselves to the case \(m=2,\) the proofs are involved and difficult, and the result generalizes a lot of proximinality results in spaces of integrable functions. The introduction contains a good survey of these results.