The Setting
Every subset of a Euclidean space is doubling, and a bi-Lipschitz image of a doubling space is again doubling. Being doubling is therefore necessary for a subset of the Hilbert space to admit a bi-Lipschitz embedding into some , and the question is whether it is also sufficient.
The library settles the question with one theorem about a single subset of , built on two definitions: a doubling subset and a bi-Lipschitz embedding.
Doubling Subsets and Bi-Lipschitz Embeddings
The Hilbert space is the real vector space of square-summable real sequences , with the distance (1)✦
Let be a natural number. A subset of is doubling with constant at most if for every point and every radius there is a set of at most points of such that (2)✦
Let be a natural number, and give its Euclidean norm. A subset of admits a bi-Lipschitz embedding into if there are a map , a scale and a distortion such that (3)✦
The Claim
Theorem 1. There is a subset of , doubling with constant at most in the sense of (2), such that for every positive integer , every map , every scale and every distortion , there are points with (4)✦
What Is Checked
The claim is stated over the Hilbert space with the distance (1), the doubling condition (2) with , and bi-Lipschitz embeddings into in the sense of (3).