Extremal random beta polytopes

From MaRDI portal



Abstract: The convex hull of several i.i.d. beta distributed random vectors in mathbbRd is called the random beta polytope. Recently, the expected values of their intrinsic volumes, number of faces, normal and tangent angles and other quantities have been calculated, explicitly and asymptotically. In this paper, we aim to investigate the asymptotic behavior of the beta polytopes with extremal intrinsic volumes. We suggest a conjecture and solve it in dimension 2. To this end, we obtain some general limit relation for a wide class of U-max statistics whose kernels include the perimeter and the area of the convex hull of the arguments.


Let \(U_1,\dots,U_N\) be independent random points in \({\mathbb R}^d\) with a beta distribution, that is, with a distribution concentrated on the unit ball \({\mathbb B}^d\) and with density proportional to \(x\mapsto (1-\|x\|^2)^\beta\), with \(\beta>-1\). Fix an integer \(n\) and let \(N>n\). The author is interested in the asymptotics (as \(N\to\infty\)) of the random variable \[ \Delta(N,n,m):=\max_{x_1,\dots,x_n\in {\mathbb B}^d} v_m([x_1,\dots,x_n])-\max_{1\leq i_1<\dots< i_n\leq N} v_m([U_{i_1},\dots,U_{i_n}]), \] where \(v_m\) is the \(m\)th intrinsic volume and \([\cdot,\dots,\cdot]\) denotes the convex hull. It is conjectured that there exist positive numbers \(A,B,C\) (depending on \(d,m,n,\beta\)) such that \[ \lim_{N\to\infty} {\mathbb P}\left[N^A \Delta(N,n,m)\leq t\right] =1-e^{-Bt^C}. \] This is proved for \(d=2\), with explicit values for \(A,B,C\). The result is derived from a more general theorem, which holds for a wider class of kernel functions for U-max statistics.











This page was built for publication: Extremal random beta polytopes

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6174444)