scientific article; zbMATH DE number 1001426

From MaRDI portal
Publication:3128708





As \(n\) grows large, the ratio between the diameters of the largest and smallest \(m\)-dimensional sections of a symmetric convex body \(K\in \mathbb{R}^n\) can grow without bound. However, one might wonder whether this behavior is typical, or whether ``most sections (in some sense) of any symmetric convex body have similar diameters. Obviously, this will not be the case for a very elongated body, say an ellipsoid with semiaxes \((l,\varepsilon, \varepsilon^2, \dots, \varepsilon^n)\). However, the problem may be redefined to compensate for the elongation of the body; or the body may always be put into a regular position in which it is ``close to spherical by a linear transformation [see \textit{G. Pisier}, `The volume of convex bodies and Banach space geometry', Cambridge (1989; Zbl 0698.46008), chapter 7]. The authors take both of these approaches. They define a decreasing function \(M^*_K: [0,\infty) \to [0,1]\) that captures certain aspects of the shape of the body \(K\), and show that there exist three functions \(f_1,f_2\), and \(g\), such that \(2g(\lambda)\) \((M^*_K)^{-1}(f_1(\lambda)) \leq\text{diam} (K\cap E) \leq 2(M^*K)^{-1} (f_2(\lambda))\) for most \([\lambda n]\)-dimensional subspaces \(E\) of \(\mathbb{R}^n\). ``Mosthere means with a set of exceptions whose measure approaches 0 exponentially as \(n\) gets large. In general, \((M^*_K)^{-1}\) may change very rapidly, so that even with \(f_1(\lambda)/f_2 (\lambda)\) close to 1, the interval may be very large. As mentioned above, this must happen in such cases. However, if the body is taken to be \(\alpha\)-regular, the authors show that there exists an interval which contains the diameters of most of the \([\lambda n]\)-dimensional sections, and such that the ratio of the upper and lower bounds depends on the shape of the body only via the regularity parameter \(\alpha\).











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3128708)