English

Superconcentration, and randomized Dvoretzky's theorem for spaces with 1-unconditional bases

Metric Geometry 2017-02-21 v3 Probability

Abstract

Let nn be a sufficiently large natural number and let BB be an origin-symmetric convex body in RnR^n in the \ell-position, and such that the normed space (Rn,B)(R^n,\|\cdot\|_B) admits a 11-unconditional basis. Then for any ε(0,1/2]\varepsilon\in(0,1/2], and for random cεlogn/log1εc\varepsilon\log n/\log\frac{1}{\varepsilon}-dimensional subspace EE distributed according to the rotation-invariant (Haar) measure, the section BEB\cap E is (1+ε)(1+\varepsilon)-Euclidean with probability close to one. This shows that the "worst-case" dependence on ε\varepsilon in the randomized Dvoretzky theorem in the \ell-position is significantly better than in John's position. It is a previously unexplored feature, which has strong connections with the concept of superconcentration introduced by S. Chatterjee. In fact, our main result follows from the next theorem: Let BB be as before and assume additionally that BB has a smooth boundary and EγnBncEγngradB()2{\mathbb E}_{\gamma_n}\|\cdot\|_B\leq n^c\,{\mathbb E}_{\gamma_n}\big\|{\rm grad}_B(\cdot)\big\|_2 for a small universal constant c>0c>0, where gradB(){\rm grad}_B(\cdot) is the gradient of B\|\cdot\|_B and γn\gamma_n is the standard Gaussian measure in RnR^n. Then for any p[1,clogn]p\in[1,c\log n] the pp-th power of the norm Bp\|\cdot\|_B^p is Clogn\frac{C}{\log n}--superconcentrated in the Gauss space.

Keywords

Cite

@article{arxiv.1702.00859,
  title  = {Superconcentration, and randomized Dvoretzky's theorem for spaces with 1-unconditional bases},
  author = {Konstantin Tikhomirov},
  journal= {arXiv preprint arXiv:1702.00859},
  year   = {2017}
}

Comments

removed the smoothness assumption in the main theorem

R2 v1 2026-06-22T18:08:10.630Z