The nearest point theorem for weakly convex sets in asymmetric seminormed spaces (Q2026144)

The paper is concerned with best approximation in asymmetric normed spaces. An asymmetric seminorm (norm) on a real vector space \(E\) is a positively homogeneous, subadditive functional \(\mu:E\to [0,\infty)\) such that \(\|x\|_\mu:=\max\{\mu(x),\mu(-x)\}>0\) (resp. \(\mu(x)>0\)) for every \(x\in E\setminus\{0\}\). In this case \(\|\cdot\|_\mu\) is a norm and \(\rho_\mu(x,y)=\mu(x-y)\), \( x,y\in E\), a quasi-metric on \(E\). The pair \((E,\mu)\) is called an asymmetric seminormed (normed) space. If the associated normed space \((E,\|\cdot\|_\mu)\) is Banach, then one says that \((E,\mu)\) is biBanach [\textit{Ş. Cobzaş}, Functional analysis in asymmetric normed spaces. Basel: Birkhäuser (2013; Zbl 1266.46001)]. For a nonempty subset \(A\) of \(E\) and \(x\in E\) define by \(\rho_\mu(x,A)=\inf\{\mu(x-a): a\in A\}\,\) the \(\mu\)-distance from \(x\) to \(A\), by \(P_\mu(x,A)=\{a\in A:\mu(x-a)=\rho_\mu(x-a)\}\,\) the \(\mu\)-projection of \(x\) on \(A\) and by \(N_\mu(a,A)=\{z\in E: \exists t>0,\, a\in A,\, a\in P_\mu(a+tz,A)\}\,\) the cone of proximal normals at \(a\in A\). The set \(A\) is called \(\mu\)-weakly convex if for all \(a\in A\) and \(z\in N_\mu(a,A)\) with \(\mu(z)=1\) one has \(a\in P_\mu(a+z,A)\). The asymmetric seminormed space \((E,\mu)\) is called parabolic if \(\sup\{\mu(-x): x\in E,\, \mu(x)\le 1,\, \mu(x+b)>2\}<\infty\) for all \(b\in E\). A notion of uniform convexity for \(\mu\), which agrees with the usual uniform convexity when \(\mu\) is a norm, is also defined. One of the main results of the paper (Theorem 2) asserts that \(P_\mu(x,A)\) is a singleton for every \(x\in E\) satisfying \(0<\rho_\mu(x,A)<1\) and for every nonempty, closed, \(\mu\)-weakly convex subset of a parabolic, uniformly convex, biBanach asymmetric seminormed space \((E,\mu)\). A result on the diameter of the set \(P_\mu^\varepsilon(x,A)\) of \(\varepsilon\)-\(\mu\)-projections is also included. An application to the Tikhonov well-posedness of a nonconvex optimization problem involving the infimal convolution of two functions is given. For the entire collection see [Zbl 1460.90004].