[Paper Review] Equilateral sets in finite-dimensional normed spaces

Konrad J. Swanepoel|ArXiv.org|Jun 14, 2004

Fixed Point Theorems Analysis30 references21 citations

TL;DR

This expository paper investigates the maximum size of equilateral sets in finite-dimensional normed spaces, particularly focusing on ℓ_p^n spaces. Using tools from linear algebra, approximation theory, and combinatorics, it establishes improved upper bounds for equilateral sets, showing e(ℓ₁ⁿ) < cn log n and e(ℓ_pⁿ) < c_p n^{(2p+2)/(2p−1)} for 1 ≤ p < ∞, resolving long-standing conjectures for specific cases like ℓ₄ⁿ and ℓ₁ⁿ.

ABSTRACT

This is an expository paper on the largest size of equilateral sets in finite-dimensional normed spaces.

Motivation & Objective

• To determine the largest possible size of equilateral sets in finite-dimensional normed spaces, especially ℓ_p^n spaces.
• To resolve Kusner’s conjectures on whether e(ℓ₁ⁿ) = 2n and e(ℓ_pⁿ) = n+1 for 1 < p < ∞.
• To improve upon the trivial 2ⁿ upper bound for e(ℓ_pⁿ) using advanced analytical and algebraic techniques.
• To establish quantitative upper bounds on equilateral set sizes using rank arguments and approximation theory.
• To explore open problems in infinite-dimensional spaces and generalizations to k-distance sets

Proposed method

• Applies the linear algebra method via rank arguments on matrices derived from equilateral sets, particularly using the rank lemma for approximations of the identity.
• Combines Smyth’s approach with Jackson’s theorems from approximation theory to bound the size of equilateral sets in ℓ_pⁿ.
• Uses randomized rounding techniques to improve bounds for ℓ₁ⁿ, leading to the bound e(ℓ₁ⁿ) < cn log n.
• Employs Cayley-Menger determinants to analyze the embeddability of almost-equilateral simplices in Euclidean space.
• Analyzes the structure of equilateral sets through geometric and topological arguments, including convex geometry and Banach space theory.
• Introduces a combinatorial framework involving intervals and hit counts to derive inequalities that constrain the size of equilateral sets

Experimental results

Research questions

• RQ1 Is e(ℓ₁ⁿ) = 2n, as conjectured by Kusner, and does e(ℓ_pⁿ) = n+1 hold for all 1 < p < ∞?
• RQ2 Can the upper bound e(ℓ_pⁿ) < 2ⁿ be improved for ℓ_pⁿ spaces with 1 < p < ∞?
• RQ3 What is the best possible upper bound for e(ℓ₁ⁿ), and does it grow as O(n log n)?
• RQ4 Can the linear algebra method combined with approximation theory yield tighter bounds for general ℓ_pⁿ?
• RQ5 Do infinite-dimensional separable Banach spaces admit infinite equilateral sets?

Key findings

• The paper proves e(ℓ₁ⁿ) < cn log n using a combination of the rank lemma and randomized rounding, resolving a key conjecture for ℓ₁ⁿ.
• For ℓ₄ⁿ, it establishes e(ℓ₄ⁿ) = n+1, confirming Kusner’s conjecture for this specific case.
• It improves the upper bound for e(ℓ_pⁿ) to c_p n^{(2p+2)/(2p−1)} for 1 ≤ p < ∞, using Smyth’s method and the rank lemma.
• For 1 < p < 2, it shows that e(ℓ_pⁿ) > n+1 when n is sufficiently large, disproving Kusner’s conjecture for this range.
• It proves that if p is sufficiently close to 2 (depending on n), then e(ℓ_pⁿ) = n+1, providing a quantitative threshold.
• The paper establishes that m < cn log n for equilateral sets in ℓ_pⁿ by deriving a contradiction from an inequality involving hit counts and interval lengths