Almost Euclidean sections in symmetric spaces and concentration of order statistics (Q2437055)

The theorem of Dvoretzky [\textit{A. Dvoretzky}, in: Proc. Int. Symp. Linear Spaces, Jerusalem 1960, 123--160 (1961; Zbl 0119.31803)] is a celebrated result in Banach space theory. It says that for every \(\varepsilon>0\) and \(m\in\mathbb N\) there exists a number \(n:=n(m,\varepsilon)\) such that any \(n\)-dimensional normed space contains an \(m\)-dimensional \((1+\varepsilon)\)-Euclidean subspace. In this paper, the author proves that for any \(n\)-dimensional normed space \(X\) with a symmetric basis and any \(\varepsilon>0\), \(X\) contains a \((1+\varepsilon)\)-Euclidean subspace of dimension at least \(c\ln(n)/\ln(\varepsilon^{-1})\), \(c\) being an absolute positive constant. This generalizes a result of \textit{Y. Gordon} et al. [Stud. Math. 178, No. 1, 91--98 (2007; Zbl 1126.46006)]. The proof is based on the concentration of order statistics of random vectors with iid coordinates.