Convergence of best approximations from unbounded sets (Q1191775)

Let \(X\) be a metric space whose bounded subsets are relatively compact (i.e. have bounded closures). For \(\emptyset\neq Y\subset X\) denote by \(P_ Y\) the operator of metric projection on \(Y\) and let \(U(Y)=\{x\in X\): \(P_ Y(x)\) is a singleton\}. Let \((S_ n)\) be a sequence of non- void subsets of \(X\) and \(S\subset X\). If \(\lim S_ n=S\), in the sense of \textit{K. Kuratowski} [Topology, Vol. I, New York (1966; Zbl 0158.408)], then for every \(x\in U(S)\) and every \(s_ n\in P_{S_ n}(x)\), the sequence \(\{s_ n\}\) converges to \(s\in P_ S(x)\) (Theorem 3.2). If \(X\) is locally compact and \(X_ \infty=X\cup\{\infty\}\) denotes its one- point Alexandrov compactification, then \(\lim s_ n=\infty\) in \(X_ \infty\) (equivalently to \(\lim S_ n=\emptyset\) in \(X\)) if and only if \(\lim d(x,S_ n)=\infty\), for all \(x\in X\) (Proposition 3.4). Section 4 is concerned with the study of convergence of linear varieties \(S_ n=\{x\in H\): \(A_ nx =y_ n\}\) to \(S=\{x\in H\): \(Ax=y\}\), for \(A_ n\) and \(A\) linear operators on a Hilbert space \(H\). Assuming \(AA^*\) invertible then \(A_ n\to A\) uniformly implies \(\lim S_ n=S\). If \(S_ n\) are closed hyperplanes in \(H\) and \(\lim S_ n=S\), then \(S\) is also a hyperplane. The obtained results are applied to approximating a solution to an infinite system of inequalities in unbounded variables and to problems of best approximation by elements of convex bodies in Euclidean space.