[Paper Review] Bi-Lipschitz approximation by finite-dimensional imbeddings

Karin U. Katz, Mikhail G. Katz|ArXiv.org|Feb 18, 2009

Topological and Geometric Data Analysis3 references4 citations

TL;DR

This paper demonstrates that any compact Riemannian manifold can be $(1+C)$-bi-Lipschitz embedded into a finite-dimensional $\ell^\infty$ space for any $C>0$, using non-standard analysis to extend the first variation formula to hyperreal points. The key contribution is a finite-dimensional approximation of the infinite-dimensional Kuratowski embedding, preserving metric structure up to arbitrary bi-Lipschitz distortion.

ABSTRACT

We show that the Kuratowski imbedding of a Riemannian manifold in L^\infty, exploited in Gromov's proof of the systolic inequality for essential manifolds, admits an approximation by a (1+C)-bi-Lipschitz (onto its image), finite-dimensional imbedding for every C>0. Our key tool is the first variation formula thought of as a real statement in first-order logic, in the context of non-standard analysis.

Motivation & Objective

• To show that the infinite-dimensional Kuratowski embedding of a compact Riemannian manifold into $L^\infty$ can be approximated by finite-dimensional embeddings.
• To resolve the misconception that infinite-dimensionality is essential in Gromov’s systolic inequality proof.
• To establish that the bi-Lipschitz distortion can be made arbitrarily close to 1 using finite-dimensional embeddings.
• To provide a rigorous foundation for finite-dimensional approximations in systolic geometry using non-standard analysis.
• To demonstrate that curvature assumptions like those in Toponogov’s theorem are not necessary, as the first variation formula suffices.

Proposed method

• Use the transfer principle from non-standard analysis to extend the first variation formula to hyperreal points.
• Apply the standard part function to map hyperreal distances back to real distances, preserving metric structure.
• Construct a finite $\epsilon$-net on the manifold to define a finite-dimensional embedding into $\ell^\infty(\mathcal{M})$.
• Leverage the injectivity radius and geodesic completeness to ensure local uniqueness of minimizing geodesics.
• Use hyperreal infinitesimals to analyze the first variation of arc length along nearby curves, extending the formula to non-standard points.
• Apply the non-standard characterization of continuity and uniform continuity to control distortion in the finite-dimensional approximation.

Experimental results

Research questions

• RQ1 Can the Kuratowski embedding of a compact Riemannian manifold into $L^\infty$ be approximated by finite-dimensional embeddings with controlled bi-Lipschitz distortion?
• RQ2 Is the first variation formula in differential geometry expressible as a first-order logical statement, enabling transfer to hyperreals?
• RQ3 Does the absence of curvature assumptions invalidate the use of Toponogov’s theorem in systolic approximation, and can the first variation formula suffice instead?
• RQ4 Can the standard part function and hyperreal extensions be used to rigorously control the distortion in finite-dimensional embeddings?
• RQ5 Is it possible to achieve $(1+C)$-bi-Lipschitz approximation for any $C>0$ using only finite-dimensional $\ell^\infty$ embeddings?

Key findings

• For every $C>0$, there exists a $(1+C)$-bi-Lipschitz finite-dimensional embedding of a compact Riemannian manifold into $\ell^\infty(\mathcal{M})$ for some finite $\mathcal{M}$.
• The finite-dimensional approximation preserves the systolic structure up to a factor of at most 5 when $\epsilon < \frac{1}{10} \mathrm{sys}(M)$, but can be improved to arbitrary distortion via finer nets.
• The first variation formula, when formulated in first-order logic, transfers to the hyperreal setting, enabling analysis of infinitesimally close points.
• The standard part function allows the recovery of real distances from hyperreal distances, ensuring the approximation remains bi-Lipschitz.
• The proof does not require curvature bounds or Toponogov’s comparison theorem, relying instead on the first variation formula and non-standard analysis.
• The transfer principle ensures that first-order statements about real geodesics and injectivity radius hold in the hyperreal extension, enabling the construction of the approximation.