Random version of Dvoretzky's theorem in \(\ell_p^n\) (Q2403703)

Dvoretzky's theorem is one of the cornerstones of asymptotic geometric analysis. In its geometric version it states that every centrally symmetric convex body in \(\mathbb R^n\) has a central section of `large' dimension which is almost spherical. More precisely, Milman's version states that: for any \(\varepsilon\in(0,1)\), there exists \(\eta(\varepsilon)>0\) such that, for every \(n\)-dimensional symmetric convex body \(C\), there exists a linear image \(\widetilde{C}\) of \(C\) and a \(k\)-dimensional subspace \(F\) with \(k\geq \eta(\varepsilon)\log(n)\) so that \[ (1-\varepsilon) B_F \subset \widetilde{C}\cap F \subset (1+\varepsilon) B_F, \] with \(B_F\) being the Euclidean ball in \(F\). In this paper, the authors study the dependence on \(\varepsilon\) in the critical dimension \(k(n,p,\varepsilon)\) for which one can find random sections of the \(\ell_p^n\)-ball which are \((1+\varepsilon)\)-spherical. The bounds obtained are shown to agree with the sharp estimates for the extreme cases \(p=1\) and \(p=\infty\). Toward this end, the authors provide tight bounds for the Gaussian concentration of the \(\ell_p\)-norm.