TheoremBase

A λ\lambda-Displacement Convex Penalty Pair Has a λ\lambda-Monotone Score Along Optimal Couplings

For a lambda-displacement convex penalty pair and two measures of the score domain, the displacement pairings of the two scores along an optimal coupling and its swap sum to at most minus lambda times the squared Wasserstein distance; at pairs uniquely mapped in both directions this reads as monotonicity of the scores against the optimal displacements.

Statement

In the setting of The Intrinsic Calculus on the Wasserstein Space: Standing Notation, let λ∈R\lambda\in\mathbb{R}, let (D,DΣ,E,Σ)(\mathcal{D},\mathcal{D}_{\Sigma},\mathcal{E},\Sigma) be a penalty pair on P2(Rd)\mathcal{P}_{2}(\mathbb{R}^{d}) that is λ\lambda-displacement convex, and let μ,ν∈DΣ\mu,\nu\in\mathcal{D}_{\Sigma}, whose scores Σ(μ)\Sigma(\mu) and Σ(ν)\Sigma(\nu) lie in L2(μ;Rd)L^{2}(\mu;\mathbb{R}^{d}) and L2(ν;Rd)L^{2}(\nu;\mathbb{R}^{d}) by Penalty Pairs on the Wasserstein Space: the Penalty, Its Score, and Their Domains (i). Displacement pairings J\mathcal{J} are those of The Intrinsic Calculus on the Wasserstein Space: Standing Notation (i), and optimal couplings those of Wasserstein Spaces, Random Vectors, Vector Fields and Symmetric Matrices in Every Dimension: Standing Notation (i). For π∈Π(μ,ν)\pi\in\Pi(\mu,\nu), π⊤∈Π(ν,μ)\pi^{\top}\in\Pi(\nu,\mu) denotes its image under the swap, which belongs to Π(ν,μ)\Pi(\nu,\mu) by that clause.

1. (λ\lambda-monotonicity along an optimal coupling) For every optimal coupling π∈Π(μ,ν)\pi\in\Pi(\mu,\nu),

λ W2(μ,ν)2 ≤ −J(Σ(μ),π)−J(Σ(ν),π⊤).\lambda\,W_{2}(\mu,\nu)^{2}\ \le\ -\mathcal{J}(\Sigma(\mu),\pi)-\mathcal{J}(\Sigma(\nu),\pi^{\top}).

2. (λ\lambda-monotonicity at uniquely mapped pairs) Suppose that both ordered pairs (μ,ν)(\mu,\nu) and (ν,μ)(\nu,\mu) are uniquely mapped, let SS be an optimal map from μ\mu to ν\nu and S′S' an optimal map from ν\nu to μ\mu, with classes id−S∈L2(μ;Rd)\mathrm{id}-S\in L^{2}(\mu;\mathbb{R}^{d}) and id−S′∈L2(ν;Rd)\mathrm{id}-S'\in L^{2}(\nu;\mathbb{R}^{d}) as in The Optimal Map as a Square-Integrable Vector Field: Integrability, Transport Cost and Uniqueness of the Class (i). Then

λ W2(μ,ν)2 ≤ ⟨Σ(μ),id−S⟩μ+⟨Σ(ν),id−S′⟩ν.\lambda\,W_{2}(\mu,\nu)^{2}\ \le\ \langle\Sigma(\mu),\mathrm{id}-S\rangle_{\mu}+\langle\Sigma(\nu),\mathrm{id}-S'\rangle_{\nu}.

Proofs

Log in to submit a proof.

Loading...

Citations

Loading…

Dependencies

Loading…

Related

0 relations

Curated associations between results. These are editable and subjective — they do not replace the dependency graph, which is derived from the references in the text.

No relations recorded yet.

Comments

Log in to comment.

Loading…