Do Minkowski averages get progressively more convex? (Q512347): Difference between revisions

For a nonempty compact set \(A\subset{\mathbb R}^n\) and for \(k\in{\mathbb N}\), let \(A(k)\) be the Minkowski sum of \(k\) copies of \(A\), scaled by \(1/k\). It is known (and strengthened by the Shapley-Folkman-Starr theorem) that the Hausdorff distance of \(A(k)\) and \(A\) tends to zero as \(k\to\infty\). This note deals with some related monotonicity questions. First, a conjecture by Bobkov, Madiman and Wang, according to which the volume of \(A(k)\) should be non-decreasing in \(k\), is disproved for \(n\geq 12\). For the non-convexity index \(c(A):=\inf\{\lambda\geq 0: A+\lambda\,\text{conv}(A)\text{ is convex}\}\), introduced by the reviewer, it is proved that \(c(A(k+1)) \leq (k/k+1)c(A(k))\). Denoting by \(d(A)\) the Hausdorff distance of \(A\) from its convex hull, it is shown that \(d(A(k+1)) \leq (k/k+1)d(A(k))\) for \(k\geq n\).