Geometric probability and random cotype 2

From MaRDI portal





The authors study properties of star bodies related to uniform (volume) distribution. In the first part, the authors consider the following multi-integral expression (actually representing a norm in \({\mathbb R}^m\)): \[ ||| \overline\lambda||| =\int_{V_1}\dots\int_{V_m} \left\| \sum_1^m\lambda_ix_i\right\| _K\frac{\prod dx_i}{\prod | V_i| }, \] where \(V_i\), \(K\) are centrally symmetric convex bodies of volume \(1\) and \({\overline\lambda}=(\lambda_i\in \mathbb R)_{i=1}^n\). This expression was introduced in [\textit{J. Bourgain, M. Meyer, V. Milman} and \textit{A. Pajor}, ``On a geometric inequality, Lect. Notes Math. 1317, 271--282 (1988; Zbl 0653.46015)]. The authors provide a new lower \(\ell_2\)-bound for \(||| \overline\lambda||| \), showing that \(||| \overline\lambda||| \geq c\sqrt{\sum\lambda_i^2}\). The authors also study the deviation from \(\ell_2\)-estimates and provide a computation for \(||| \overline\lambda||| \) in the case of \(V_i=K=[-1,1]^n\) (the unit ball of \(\ell_\infty^n\)), showing that in this case \(||| \overline\lambda||| \) is very (actually \(\log n\)) close to the \(\ell_2\)-norm. Next, they introduce the notion of random cotype 2 and show that any finite-dimensional normed space has random cotype 2. Finally, the authors discuss a possibility to check in the ``reasonable time that a finite-dimensional normed space is very far from being Euclidean. For the entire collection see [Zbl 1052.46001].











This page was built for publication: Geometric probability and random cotype 2

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4664563)