OctoInformBeta

OpenAI preprint, September 2026 · §7. Compact obstructions in infinite-dimensional Banach spaces

Finite-dimensional targets: A bi-Lipschitz map from a metric space into an nn-dimensional real normed space yields one into Euclidean ℝn\mathbb{R}^n, with other constants

A Euclidean target suffices for finite-dimensional bi-Lipschitz embeddings.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Take a metric space (X,d)(X,d) and a bi-Lipschitz map of it into a real normed space EE of finite dimension nn, such as ℝn\mathbb{R}^n with the maximum norm. The question is whether the norm of the target matters: does XX then admit a bi-Lipschitz map into ℝn\mathbb{R}^n with the Euclidean norm?

A bi-Lipschitz embedding of a metric space (X,d)(X,d) into a real normed space EE at scale a>0a>0 with distortion D≥1D\ge 1 is a map f:X→Ef\colon X\to E such that ad(x,y)≤∥f(x)−f(y)∥≤Dad(x,y)for all x,y∈X.a\,d(x,y)\le \|f(x)-f(y)\|\le D\,a\,d(x,y)\qquad\text{for all } x,y\in X.(1)

The Claim

Theorem 1. Let (X,d)(X,d) be a metric space, EE a real normed space of finite dimension nn, and ff a bi-Lipschitz embedding of XX into EE at scale a>0a>0 with distortion D≥1D\ge 1. Then there are a map g:X→ℝng\colon X\to\mathbb{R}^n, where ℝn\mathbb{R}^n carries the Euclidean norm |⋅||\cdot|, a scale a′>0a'>0 and a distortion D′≥1D'\ge 1 such that a′d(x,y)≤|g(x)−g(y)|≤D′a′d(x,y)for all x,y∈X.a'\,d(x,y)\le |g(x)-g(y)|\le D'\,a'\,d(x,y)\qquad\text{for all } x,y\in X.(2)✦

What Is Checked

The claim (2) is stated over an arbitrary metric space XX, a real normed space EE of finite dimension nn and a map f:X→Ef\colon X\to E satisfying (1) for some a>0a>0 and D≥1D\ge 1, and the map gg it gives takes values in ℝn\mathbb{R}^n with the Euclidean norm.