On separation by families of linear functionals (Q1174991): Difference between revisions

Let \(S\) be a convex subset of a Banach space \(L\) and \(G\) a subset of \(L^*\). For each positive integer \(n\), write \(G_ n\) for the collection of all linear functionals which can be expressed as linear combinations of \(n\) or fewer elements of \(G\) and let \(G_ \infty\) denote the full linear span of \(G\). \(G\) is said to separate two subsets \(A\), \(B\) of \(L\) if there is a \(g\in G\) for which \(g(A)\) and \(g(B)\) have disjoint closures. For \(G\subseteq H\subseteq L^*\) and \(S\) a convex subset of \(L\), define ``\(G\sim H\) in \(S\)'' if \(G\) and \(H\) are equally effective at separating points in \(S\) from convex subsets of \(S\) and ``\(G\sim H\) outside \(S\)'' if \(G\) and \(H\) are equally effective at separating points outside \(S\) from \(S\). Then the main results proved by the author are given below: Theorem: If the linear span of either \(S\) or \(G\) is finite-dimensional then \(G\sim G_ \infty\) in \(S\) implies \(G_ 3\sim G_ \infty\) outside \(S\). Corollary: Let \({\mathcal G}\) be a family of planes in \(R\), each triple of which intersects in the first octant. If the normal vectors of \({\mathcal G}\) form a dense subset of \((R^ 3)^*\), then \({\mathcal G}\) is concurrent. The theorem above is sharp as \(G_ 3\) cannot be replaced by \(G_ 2\) and the two-dimensional analogue of the corollary fails. There are many other interesting results in this paper.