Best approximation by normal and conormal sets (Q5929848): Difference between revisions

Let \(\mathbb R_+^n\) be the set of all elements in \(\mathbb R^n\) with non-negative coordinates. A set \(G\subseteq \mathbb R_+^n\) is called normal (conormal) if \(g\in G\), and \(0\leq x\leq g\) implies \(x\in G, (\mathbb R_+ ^n\setminus G\) is normal\(),\) respectively. The aim of this paper is to develop a theory of best approximation in \((\mathbb R^n, \|\cdot\|_\infty)\) by elements in a normal (conormal) set. An important tool in this study is the family of min-type functions as analogous of linear functionals. For each \(\ell\in\mathbb R_+^n, \ell\neq 0\) the function \(<\ell,\cdot>:\mathbb R_+^n\rightarrow \mathbb R\) given by \(<\ell ,x>=\min\{\ell_ix_i:i\in\{ 1,...,n\}, {\text{with}} \ell_i>0\}, x\in \mathbb R^n_+,\) is called a min-type function. A characterization of normal or closed normal subsets of \(\mathbb R_+^n\) is given in terms of min-type functions. As in the approximation by elements in convex sets a normal set \(G\) and the ball \(B(x_0,d(x_0,G)), x_0\in \mathbb R_+^n\setminus G\) can be separated, this time by min-type hyperplanes. In contrast to the convex case the least element of best approximation \(g_0\) of \(x_0\) plays an important role, especially when \(g_0\) is a ''weak Pareto point'' of \(G.\) Characterizations of nearest points in normal (conormal) sets are given and the distance to a min-type hyperplane is computed.