The randomized integer convex hull

From MaRDI portal





Let \({\mathbb Z}^d\) denote the \(d\)-dimensional integer lattice. Let \(K\subset {\mathbb R}^d\) be a convex body, i.e. a convex compact set with nonempty interior. The integer convex hull \(I(K)\) of \(K\) is the convex hull of all lattice points in \(K\), i.e. \(I(K) = conv(K\cap {\mathbb Z}^d)\). \(I(K)\) is a convex polytope which is of central interest in integer programming. For instance, when maximizing a linear function \(f\) on the integer points in \(K\), one looks for the maximum of \(f\) on \(I(K)\). The authors consider a randomized version of the integer convex hull and expected missed area. The lattice \({\mathbb Z}^d\) is replaced by \(L\), a randomly translated and rotated copy of \({\mathbb Z}^d\), i.e. \(I_L(K) = conv(K\cap L)\). More precisely, for a translation vector \(t\in [0,1)^d\) and a rotation \(\rho\in SO(d)\) around the origin, we set \(L_{t,\rho}= \rho\,({\mathbb Z}^d+t)\), and define \({\mathcal L}\)= \(\{L_{t, \rho}:t\in [0,1)^d, \rho \in SO(d)\}\). A probability measure on \({\mathcal L}\) is defined as the product of Haar measures on \([0,1)^d\) and \(SO(d)\) respectively. The article is organized as follows. In section 2 properties of the Macbeath regions are given. In section 3, the upper and lower bounds for the expected number of vertices in terms of the volume of the so-called Macbeath region are obtained. In section 4, in the planar case, the similar problem for expected missed area is investigated. The last section contains some open problems.











This page was built for publication: The randomized integer convex hull

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1764169)