Hadamard products and moments of random vectors (Q2213794): Difference between revisions

The authors establish an upper bound on strong moments of finite dimensional random vectors, in terms of the corresponding weak moments. In particular, for a random vector \(X\) on a normed space \((\mathbb{R}^n,\lVert\cdot\rVert)\), one of the main results of the paper is the upper bound \[ (\mathbb{E}\lVert X^p\rVert)^{1/p}\leq2\sqrt{e}\sqrt{\frac{n+p}{p}}\sup_{\lVert t\rVert_*\leq1}(\mathbb{E}|\langle t,X\rangle|^p)^{1/p}\,, \] for any \(p\geq2\), where \(\lVert\cdot\rVert_*\) denotes the dual norm. Two applications are discussed: the authors show that (up to a universal constant) Hilbert spaces have the largest \(p\)-summing constant among Banach spaces of a fixed dimension, and establish concentration of measure properties for centered log-concave probability measures on \(\mathbb{R}^n\).