Uniformly and locally convex asymmetric spaces (Q2120157)

The author continues his investigation of uniform and local uniform convexity in asymmetric normed spaces initiated in [\textit{I. G. Tsar'kov}, Math. Notes 110, No. 5, 773--783 (2021; Zbl 1489.46022); translation from Mat. Zametki 110, No. 5, 773--785 (2021)]. The definitions of these notions are given in such a way that many properties of uniformly convex and locally uniformly convex (symmetric) normed spaces are preserved in the asymmetric case. An asymmetric norm on a real vector space \(X\) is a subadditive, positively homogeneous functional \(\|\cdot|:X\to[0,\infty)\) such that \(\|x|=0\iff x=0\). If the weaker condition \(\|x|=0=\|{-x}|\;\Rightarrow\; x=0\) holds, then \(\|\cdot|\) is called an asymmetric seminorm. Note that an asymmetric seminormed space satisfies only the \(T_0\) separation axiom, while an asymmetric normed space is \(T_1\) (not necessarily \(T_2\)). Concerning the theory of asymmetric normed spaces see the book [\textit{S. Cobzaş}, Functional analysis in asymmetric normed spaces. Basel: Birkhäuser (2013; Zbl 1266.46001)]. A sequence \(\{x_n\}\) in \(X\) is called fundamental (inversely fundamental) if for every \(\varepsilon>0\) there exists \(N\in\mathbb{N}\) such that \(\|x_m-x_n|<\varepsilon\) (resp. \(\|x_n-x_m|<\varepsilon\)) for all \(N\le n\le m\). The space \(X\) is called right (left) complete if for every fundamental sequence \(\{x_n\}\) in \(X\) there exists \(x\in X\) such that \(\|x-x_n|\to 0\) (resp. \(\|x_n-x|\to 0\)). Similar definitions are given for inverse completeness. (Notice that the terminology of this paper slightly differs from that in the mentioned book.) The aim of the present paper is to study some best approximation problems (uniqueness, existence, and stability of elements of best approximation) as well as the nonemptiness of the intersection of a nested family of closed convex sets. For a nonempty subset \(M\) of \(X\) and \(x\in X\) one denotes by \(\rho(x,M)=\inf\{\|y-x|:y\in M\}\) the distance from \(x\) to \(M\) and by \(P_Mx=\{y\in M:\|y-x|=\rho(x,M)\}\) the set of nearest points to \(x\) in \(M.\) If, for every \(x\in M\), \(P_Mx\ne\emptyset\), then the set \(M\) is called proximinal and Chebyshev if \(P_Mx\) is a singleton. For instance, if \(x^*\) is a bounded linear functional on \(X\) of unit norm, then the set \(\{y\in X:x^*(y)\ge 1\}\) is proximinal if and only if \(x^*\) attains its norm on the unit ball of \(X\). In a left-complete asymmetric normed space \(X\), the distance function \(\rho(\cdot,M)\) is lower semicontinuous and continuous on a non-meager set (this also happens for any bounded linear functional on \(X\)). Also, every nonempty closed convex subset \(M\) of a uniformly convex right inversely complete asymmetric normed space \(X\) is Chebyshev and, for every \(x\in X\), every minimizing sequence \(\{y_n\}\) in \(M\) (i.e., such that \(\lim_{n\to\infty}\|y_n-x|=\rho(x,M))\) satisfies the condition \(\lim_{n\to\infty}\|y_n-y|=0,\) where \(y\) is the unique nearest point to \(x\) in \(M\). Under the same hypotheses, the metric projection \(P_M\) is continuous at every point where \(\rho(\cdot,M)\) is continuous.