Best approximation in certain classes of normed linear spaces (Q1060374): Difference between revisions

Let E be a (real) normed linear space. For each \(x\in E\), let \(N(x)=\{y\in E\); \(\tau (x,y)+\tau (x,-y)=0\}\) and \(A(x)=\{f\in E^*\); \(f(x)=\| x\|\), \(\| f\| =1)\), where \(\tau\) (x,y) denotes the one-sided Gateaux differential of the norm at x in the direction y, \(E^*\) denotes the conjugate space of E. Beforehand, the author of this paper introduced two classes of normed linear space called ''with property (C)'' and ''with property (\(\Lambda)\)'' as follows. The space E is called with property (C) if for each \(x\in S_ E\) (the unit sphere of E) and \(a\in A(x)_{\perp}\cap S_ E=\{x\in S_ E\); \(f(x)=0\) for each \(f\in A(x)\}\), we have \(a=(z_ 1-z_ 2)/2\) for some \(z_ i\in S_ E\) with \(A(x)\subset A(z_ i)\), \(i=1,2\). The space E is called with property (\(\Lambda)\) if \(\tau (x,y)+\tau (x,-y)=2 dist(y,N(x))\) for each x, \(y\in S_ E.\) In this paper he well applies them to the theory of best approximation and mainly proves following theorems: (i) Let E be a space with property (C) and G a sun of E. Then G is semi-Chebyshev if and only if for each \(x\in E\setminus G\) and each \(g_ 0\in P_ G(x)\) (the set of all best approximations of x out of G) we have \((G-g_ 0)\cap A(x-g_ 0)_{\perp}=0\). (ii) Let E be a space with property (C) and (\(\Lambda)\) such that for each \(x\in E\), N(x) is proximinal, and let G be a subset of E. Then the following assertions are equivalent: (1) G is a semi-Chebyshev sun; (2) For each \(x\in E\setminus \bar G\) and \(g_ 0\in P_ G(x)\), the pair \((g_ 0,x)\) satisfies the strict Kolmogorov criterion, i.e., \(\tau (x-g_ 0,g_ 0-g)>0\) for each \(g\in G\setminus \{g_ 0\}\). Since the author has discussed tha some classical spaces belong to one or both of the two classes as is stated above, and has remarked that for some spaces, N(x) is proximinal for each element x, results proved in this paper contain some previous results in fact.