OctoInformBeta

Sharp one-third stability of Brenier maps

Fix an integer d≥2d\ge2, a compact convex set K⊂ℝdK\subset\mathbb R^d with nonempty interior, and a nonempty compact set Y⊂ℝdY\subset\mathbb R^d. Let ρ\rho be normalized Lebesgue measure on KK. Write 𝒫(Y)\mathcal P(Y) for the Borel probability measures supported on YY. For μ∈𝒫(Y)\mu\in\mathcal P(Y), let TμT_\mu denote the quadratic optimal transport map from ρ\rho to μ\mu, up to a ρ\rho-null set. Write W2W_2 for the quadratic Wasserstein distance. The stability question concerns uniform control of ∥Tμ−Tν∥L2(ρ)\|T_\mu-T_\nu\|_{L^2(\rho)} in terms of W2(μ,ν)W_2(\mu,\nu) for μ,ν∈𝒫(Y)\mu,\nu\in\mathcal P(Y).

Brenier’s theorem identifies these optimal maps with gradients of convex potentials. Divol, Niles-Weed, and Pooladian proved a W21/3W_2^{1/3} estimate under regularity and nondegeneracy hypotheses. Letrouit conjectured uniform square-root stability in W2W_2 for uniform measures on convex bodies.

The preprint proves that there is a finite constant C(K,Y)≥0C(K,Y)\ge0 such that ∥Tμ−Tν∥L2(ρ)≤C(K,Y)W2(μ,ν)1/3\|T_\mu-T_\nu\|_{L^2(\rho)}\le C(K,Y)W_2(\mu,\nu)^{1/3} for all μ,ν∈𝒫(Y)\mu,\nu\in\mathcal P(Y). Every exponent α>1/3\alpha>1/3 fails uniformly already for K=Y=[−1,1]dK=Y=[-1,1]^d and targets with exactly three atoms. The preprint also gives an explicit admissible stability constant that the Lean library does not establish.

Lean has checked every result of the Lean library, so the Lean library is the reference for the statements on these pages. Each result is stated in the Lean library’s form and linked to the corresponding result in the Lean library, with any differing preprint statement beside it. The preprint is a guide only to the order and names of the results.

The Lean library: Brenier

Brenier is a Lean library: 16 items, in six groups. Each has a page written by OctoInform, listed below.

§1. Introduction 2 items
  1. §1 (setting): uniform source, target laws and quadratic optimal maps Uniform sources and quadratic transport maps.
  2. Theorem 1.1 Brenier maps have sharp one-third stability.
§2. Sharpness on a fixed cube 3 items
  1. §2 (setting): affine maxima and the three-atom cube family Affine maxima on the fixed cube.
  2. Lemma 2.1 Measurable affine maximizers and unique optimal transport.
  3. Proposition 2.2 The one-third exponent is sharp on cubes.
§3. Finite affine potentials and prescribed cell masses 3 items
  1. §3 (setting): Laguerre cells, masses, moments and face coefficients Cell masses and canonical face coefficients.
  2. Proposition 3.1 Cell variations and the canonical Laplacian.
  3. Proposition 3.2 Prescribed positive masses determine normalized intercepts.
§4. A uniform estimate for cell moments 2 items
  1. Lemma 4.1 A uniform estimate for cell moment derivatives.
  2. Lemma 4.2 Couplings controlling the midpoint law of convex bodies.
§5. Moving sites with fixed masses 2 items
  1. §5 (setting): optimal atomic motions and normalized intercept paths Optimal atomic motions and their intercept paths.
  2. Proposition 5.1 Fixed masses control differences of potentials.
§6. From potentials to maps 4 items
  1. §6 (setting): bounded supporting maps and coordinate slices Directional bounds for globally supporting maps.
  2. Lemma 6.1 Supporting maps obey a gradient interpolation bound.
  3. Corollary 6.2 Uniform constants in one-third transport stability.
  4. Proposition 6.3 Optimal maps for compact target laws.

No item matches.