Assouad's theorem with dimension independent of the snowflaking (Q1759761): Difference between revisions

The main purpose of the paper is to show that the optimal dimension for the Assouad embedding [\textit{P. Assouad}, Bull. Soc. Math. Fr. 111, No. 4, 429--448 (1983; Zbl 0597.54015)] depends on the doubling constant only and not on the snowflaking. More precisely, the main result is: There exists an absolute constant \(N\in(0,\infty)\) such that for any \(K\in(0,\infty)\), any separable metric space \((X,d)\) with doubling constant \(K\), and any \(\alpha\in(\frac12,1)\), the metric space \((X,d^\alpha)\) admits a bilipschitz embedding into \(\mathbb{R}^n\) with \(n=\lceil N\log K\rceil\) and with distortion \(\leq N(\log K/(1-\alpha))^2\). The paper contains even more precise estimates. The proof uses random embeddings and the Lovász Local Lemma. The paper also contains interesting remarks on the (open) Lang-Plaut problem [\textit{U. Lang} and \textit{C. Plaut}, Geom. Dedicata 87, No. 1--3, 285--307 (2001; Zbl 1024.54013)]: Does every doubling subset of a Hilbert space admit a bilipschitz embedding into \(\mathbb{R}^n\) for sufficiently large \(n\in\mathbb{N}\)? The paper concludes with a related result on embeddings of the Heisenberg group. Reviewer's remark. Recently, \textit{G. David} and \textit{M. Snipes} [Anal. Geom. Metr. Spaces, 36--41 (2013; Zbl 1261.53039)] obtained a simpler non-probabilistic proof of the authors' result on the independence of the optimal dimension for the Assouad embedding on the snowflaking, but their estimates for \(n\) and the distortion are weaker.