TheoremBase

The Coordinate Fields of a Coupling, and the Second Moment as a Lipschitz Function of the Wasserstein Distance

For a coupling of two probability measures with finite second moment, the two coordinate projections are square-integrable vector fields against the coupling whose squared norms are the two second moments and whose squared distance is the quadratic cost; consequently the square root of the second moment is Lipschitz with constant one for the Wasserstein distance, and the second moment is bounded on every Wasserstein ball.

Statement

In the setting of Wasserstein Spaces, Random Vectors, Vector Fields and Symmetric Matrices in Every Dimension: Standing Notation, let m∈Nm\in\mathbb{N} satisfy 1≤m1\le m, and let μ,ν∈P2(Rm)\mu,\nu\in\mathcal{P}_{2}(\mathbb{R}^{m}), the second moment M2M_{2}, the Wasserstein space P2(Rm)\mathcal{P}_{2}(\mathbb{R}^{m}), the couplings Π(μ,ν)\Pi(\mu,\nu) with their quadratic cost II, the distance W2W_{2} and the spaces L2(ρ;Rm)L^{2}(\rho;\mathbb{R}^{m}) being those fixed in Wasserstein Spaces, Random Vectors, Vector Fields and Symmetric Matrices in Every Dimension: Standing Notation (i) and Wasserstein Spaces, Random Vectors, Vector Fields and Symmetric Matrices in Every Dimension: Standing Notation (ii). Write pr1m,pr2m:Rm+m→Rm\mathrm{pr}_{1}^{m},\mathrm{pr}_{2}^{m}:\mathbb{R}^{m+m}\to\mathbb{R}^{m} for the coordinate projections of Rm+m\mathbb{R}^{m+m}. Then the following hold.

1. (The coordinate fields of a coupling) Let π∈Π(μ,ν)\pi\in\Pi(\mu,\nu). Then the classes of pr1m\mathrm{pr}_{1}^{m} and pr2m\mathrm{pr}_{2}^{m} belong to L2(π;Rm)L^{2}(\pi;\mathbb{R}^{m}), and

∥pr1m∥π2=M2(μ),∥pr2m∥π2=M2(ν),∥pr1m−pr2m∥π2=I(π).\lVert\mathrm{pr}_{1}^{m}\rVert_{\pi}^{2}=M_{2}(\mu),\qquad\lVert\mathrm{pr}_{2}^{m}\rVert_{\pi}^{2}=M_{2}(\nu),\qquad\lVert\mathrm{pr}_{1}^{m}-\mathrm{pr}_{2}^{m}\rVert_{\pi}^{2}=I(\pi).

2. (Lipschitz bound)

∣M2(μ)−M2(ν)∣≤W2(μ,ν).\Bigl|\sqrt{M_{2}(\mu)}-\sqrt{M_{2}(\nu)}\Bigr|\le W_{2}(\mu,\nu).

3. (Bound on a Wasserstein ball) If R∈RR\in\mathbb{R} is nonnegative and W2(μ,ν)≤RW_{2}(\mu,\nu)\le R, then

M2(ν)≤(M2(μ)+R)2.M_{2}(\nu)\le\Bigl(\sqrt{M_{2}(\mu)}+R\Bigr)^{2}.

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…