[Paper Review] Embedding proper ultrametric spaces into $\ell_p$ and its application to nonlinear Dvoretzky's theorem

Kei Funano|arXiv (Cornell University)|Mar 8, 2012

Advanced Topology and Set Theory10 references3 citations

TL;DR

This paper proves that every proper ultrametric space admits an isometric embedding into $\ell_p$ for any $p \geq 1$, using a recursive partitioning and coordinate assignment method based on nested closed balls. As an application, it establishes an $\ell_p$-version of nonlinear Dvoretzky's theorem, showing that every compact metric space contains a subset of controlled Hausdorff dimension that embeds into $\ell_p$ with distortion arbitrarily close to 2.

ABSTRACT

We prove that every proper ultrametric space isometrically embeds into $\ell_p$ for any $p\geq 1$. As an application we discuss an $\ell_p$-version of nonlinear Dvoretzky's theorem.

Motivation & Objective

• To resolve Lemin’s problem on isometric embedding of separable ultrametric spaces into $\ell_p$ by focusing on the subclass of proper ultrametric spaces.
• To establish a new method for constructing isometric embeddings into $\ell_p$ spaces using hierarchical partitioning of compact ultrametric spaces.
• To apply the embedding result to derive a nonlinear Dvoretzky-type theorem in $\ell_p$-spaces, extending prior results in Hilbert and $\ell_1$-settings.

Proposed method

• Constructing a nested family of partitions of a compact ultrametric space into closed balls of strictly decreasing radius.
• Assigning coordinates in $\ell_p$ to selected points in each ball using differences of $p$-th powers of radii, ensuring orthogonality and isometry.
• Extending the embedding from a countable dense subset to the full space via completeness and density arguments.
• Proving that the resulting map preserves all distances exactly by construction, relying on the ultrametric triangle inequality.
• Applying the main embedding result to Mendel and Naor’s nonlinear Dvoretzky theorem to derive an $\ell_p$-version with distortion control.
• Using random graph-based metric spaces to show the sharpness of distortion bounds, demonstrating that distortion less than 2 forces zero Hausdorff dimension on subsets.

Experimental results

Research questions

• RQ1 Can every proper ultrametric space be isometrically embedded into $\ell_p$ for any $p \geq 1$?
• RQ2 Does the existence of such embeddings allow for a nonlinear Dvoretzky-type theorem in $\ell_p$-spaces?
• RQ3 Is the distortion bound of 2 in the nonlinear Dvoretzky theorem sharp for $\ell_p$-embeddings, or can it be improved?
• RQ4 What is the relationship between the Hausdorff dimension of a metric space and the dimension of its subsets that embed into $\ell_p$ with small distortion?
• RQ5 Can the impossibility result for distortion less than 2 be extended to $\ell_p$-spaces using random graph metrics?

Key findings

• Every proper ultrametric space admits an isometric embedding into $\ell_p$ for any $p \geq 1$, resolving Lemin’s problem for this class.
• The embedding is constructed via a recursive partitioning of the space into nested closed balls and assigning orthogonal coordinates in $\ell_p$ based on radius differences.
• The method yields an isometric embedding into $c_0$ as a byproduct, differing from prior constructions.
• An $\ell_p$-version of nonlinear Dvoretzky’s theorem is established: every compact metric space contains a closed subset $S$ with $\dim_H(S) \geq \frac{c\varepsilon}{\log(1/\varepsilon)}\dim_H(X)$ that embeds into $\ell_p$ with distortion $2+\varepsilon$.
• For any $p \geq 1$, there exist compact metric spaces where any subset embedding into $\ell_p$ with distortion strictly less than 2 must have zero Hausdorff dimension.
• The distortion bound of 2 is sharp in the sense that improving it below 2 forces the embedded subset to be negligible in Hausdorff dimension.