An extension of a Bourgain-Lindenstrauss-Milman inequality (Q2461222)

Let \(\| \cdot \| \) be a norm on \(\mathbb R^n\). Consider the following unconditional norm on \(\mathbb R^n\) \[ | | | (x_1, \dots, x_n)| | | := \text{Ave} \biggl\| \sum_{i=1}^n \varepsilon_i x_i e_i\biggr\|, \] where the average is taken over all \(2^n\) choices of signs \(\varepsilon_i = \pm 1\). \textit{J.~Bourgain, J.~Lindenstrauss} and \textit{V. D.~Milman} [Lect. Notes Math. 1317, 44--66 (1988; Zbl 0645.52001)] showed that there exists an absolute constant \(C>1\) such that it is enough to take the average over \(N:= C n\) (random) choices of vectors \((\varepsilon _1, \dots, \varepsilon _n)\) in order to obtain a norm isomorphic to \(| | | \cdot| | | \). In the present paper, the authors improve the result to the case of an arbitrary \(N>n\) with constants of isomorphism depending only on \((N-n)/n\).