A probabilistic take on isoperimetric-type inequalities (Q436241): Difference between revisions

In [Arch. Math. 25, 86--90 (1974; Zbl 0287.52009)], \textit{H. Groemer} proved that the minimum of the functional giving the expected volume of a random polytope \(\text{conv}\{x_1,\dots,x_N\}\) in a convex body \(K\), i.e., \[ \Phi(K)=\frac{1}{\text{vol}_n(K)}\int_K\dots\int_K\text{vol}_n\bigl(\text{conv}\{x_1,\dots,x_N\}\bigr)dx_1\dots dx_N, \] is attained by ellipsoids. In the paper under review, the authors extend the above result in several ways. For instance, they work in the class \(\mathcal{P}_{[n]}\) of all absolutely continuous (w.r.t. the Lebesgue measure) probability measures on \(\mathbb{R}^n\), and show that if \(\mu_1,\dots,\mu_N\in\mathcal{P}_{[n]}\), \(N\geq n\), with densities \(f_1,\dots,f_N\) such that \(\|f_i\|_{\infty}\leq 1\) for \(i=1,\dots,N\), and \(C\) is an \(N\)-dimensional convex body, then the functional \[ \mathcal{F}_C(f_1,\dots,f_N)=\int_{\mathbb{R}^n}\dots\int_{\mathbb{R}^n} \text{vol}_n\bigl([x_1,\dots,x_N]C\bigr)\prod_{i=1}^Nf_i(x_i)\,dx_N\dots dx_1, \] is minimized when \(f_i=\mathbf{1}_{D_n}\) for all \(i=1,\dots,N\); here \(D_n\) denotes the Euclidean unit ball of volume one, and \([x_1,\dots,x_N]C\) is the random convex set in \(\mathbb{R}^n\) which is obtained applying on \(C\) the \(n\times N\) random matrix \([x_1,\dots,x_N]\). As consequences of these extensions, the authors get, among others, results which were previously proved by \textit{J. Bourgain} et al. in [Lect. Notes Math. 1317, 271--282 (1988; Zbl 0653.46015)] or by \textit{E. Lutwak, D. Yang} and \textit{G. Zhang} in [J. Differ. Geom. 84, No. 2, 365--387 (2010; Zbl 1206.49050)].