Randomized isoperimetric inequalities (Q2406329)

Many important geometric inequalities, such as the isoperimetric or the Brunn-Minkowski inequalities, admit stronger empirical versions associated with random convex sets. For instance, one form of the Brunn-Minkowski inequality states that if \(K\), \(L\) are convex bodies and \(B\) is the Euclidean unit ball, then \(V(K+L)\geq V(r_KB+r_LB)\), where \(r_K=(V(K)/V(B))^{1/n}\) is the radius of the Euclidean ball satisfying \(V(r_KB)=V(K)\). Now, denoting by \(K_N\) the random polytope \(K_N=\mathrm{conv}\{x_1,\dots,x_N\}\) given by the convex hull of \(N\) independent random vectors \(x_i\) distributed according to the uniform density on \(K\), \(f_K=(1/V(K))1_K\), then the expectation \[ \mathbb E V(K_{N_1}+L_{N_2})\geq \mathbb EV((r_KB)_{N_1}+(r_LB)_{N_2}). \] The above version of the Brunn-Minkowski inequality can be obtained when \(N_1\), \(N_2\to\infty\). The paper under review provides a far-reaching overview of related stochastic forms of isoperimetric inequalities. For the entire collection see [Zbl 1377.52002].