Continuity points and continuous selections of the set-valued metric generalized inverse in Banach spaces (Q2279932)

For a subspace \(Z\) of a normed space \(E\), denote by \(P_Z\) the metric projection of \(E\) onto \(Z\). The subspace \(Z\) is called proximinal, Chebyshev, or \(k\)-Chebyshev if for every \(x\in E\), \(P_Z(x)\) is nonempty, a singleton or the affine dimension of the set \(x-P_Z(x)\) is \(\le k\), respectively. Let \(X,Y\) be Banach spaces and \(T:X\to Y\) a bounded linear operator with domain \(D(T)\subset X,\) range \(R(T)=\{Tx:x\in D(T)\}\subset Y\) and null space \(N(T)=\{x\in D(T): Tx=0\}\). For \( y\in Y\), a best approximation solution (b.a.s.) to the equation \(Tx=y\) is an element \(x_0\in X \) such that \(Tx_0\in P_{R(T)}(y)\) and \(\|x_0\|=\inf \{\|v\|: Tv\in P_{R(T)}(y)\}.\) The metric generalized inverse of \(T\) is the set-valued operator \(T^\partial:D(T^\partial)\rightrightarrows Y\), where \(T^\partial(y)\) is the set of all \(x_0\in D(T)\) which are b.a.s. to the equation \(Tx=y\), for \(y\in D(T^\partial):=\{y\in Y : Tx= y\) has a b.a.s\(\}.\) One also considers the function \(g^\partial :Y\to\mathbb{R}\) given by \(g^\partial(y)=\inf\{\|v\|:Tv\in P_{R(T)}(y)\}\). It turns out that the continuity properties of the set-valued operator \(T^\partial\) are tightly connected with those of the function \(g^\partial.\) For instance, if \(D(T)\) is a closed subspace of \(X\), \(N(T)\) a Chebyshev subspace of \(D(T)\), and \(R(T)\) a 2-Chebyshev subspace of \(Y\), then \(T^\partial \) is usc at \(y_0\) iff \(g^\partial\) is continuous at \(y_0\) (see [the authors, J. Math. Anal. Appl. 422, No. 2, 1363--1375 (2015; Zbl 1311.46017)]). In this paper, the authors continue their investigations on the continuity and selection properties of the generalized inverse operator in connection with the geometric and approximation properties of the spaces \(N(T)\) in \(D(T)\) and \(D(T)\) in \(Y\), and the continuity properties of the corresponding metric projections. For instance, if \(D(T)\) is a closed subspace of \(X\), \(N(T)\) is a 2-Chebyshev subspace of \(D(T)\), and \(R(T)\) a 2-Chebyshev subspace of \(Y\), then \(T^\partial \) is lsc at \(y_0\) iff for every \(z\in T^\partial(y_0)\) there exists a selection \(T^\sigma\) of \(T^\partial\), continuous at \(y_0\) and such that \(T^\sigma(y_0)=z\) (Theorem 3.1).