TheoremBase

The Wasserstein Distance on a Hilbert Space as the Increasing Limit of the Euclidean Wasserstein Distances of the Coordinate Projections

The Wasserstein distance on a Hilbert space is the increasing limit of the Euclidean Wasserstein distances between the laws of the first n coordinates, and projecting onto the first n coordinates costs at most the tail second moment.

Statement

In the setting of Borel Probability Measures on a Real Hilbert Space with an Orthonormal Basis: Standing Notation, with the maps pn,Pn,Qnp_{n},P_{n},Q_{n} of Borel Probability Measures on a Real Hilbert Space with an Orthonormal Basis: Standing Notation §coordinates, let P2(X)\mathcal{P}_{2}(X) be the set of probability measures on XX with finite second moment and W2W_{2} the quadratic Wasserstein distance on it. For n∈Nn\in\mathbb{N}, let P2(Rn)\mathcal{P}_{2}(\mathbb{R}^{n}) be the set of probability measures on Rn\mathbb{R}^{n} with finite second moment and W2(n)W_{2}^{(n)} the quadratic Wasserstein distance on it. Let μ,ν∈P2(X)\mu,\nu\in\mathcal{P}_{2}(X) and n∈Nn\in\mathbb{N}.

1. (Projected measures) (pn)#μ∈P2(Rn)(p_{n})_{\#}\mu\in\mathcal{P}_{2}(\mathbb{R}^{n}) and (Pn)#μ∈P2(X)(P_{n})_{\#}\mu\in\mathcal{P}_{2}(X).

2. (Isometry) W2((Pn)#μ,(Pn)#ν)=W2(n)((pn)#μ,(pn)#ν)W_{2}\bigl((P_{n})_{\#}\mu,(P_{n})_{\#}\nu\bigr)=W_{2}^{(n)}\bigl((p_{n})_{\#}\mu,(p_{n})_{\#}\nu\bigr).

3. (Monotonicity) W2(n)((pn)#μ,(pn)#ν)≤W2(n+1)((pn+1)#μ,(pn+1)#ν)≤W2(μ,ν)W_{2}^{(n)}\bigl((p_{n})_{\#}\mu,(p_{n})_{\#}\nu\bigr)\le W_{2}^{(n+1)}\bigl((p_{n+1})_{\#}\mu,(p_{n+1})_{\#}\nu\bigr)\le W_{2}(\mu,\nu).

4. (Tails) W2((Pn)#μ,μ)2≤∫X∣Qnx∣2 μ(dx)W_{2}\bigl((P_{n})_{\#}\mu,\mu\bigr)^{2}\le\int_{X}|Q_{n}x|^{2}\,\mu(dx), and the real sequence (∫X∣Qmx∣2 μ(dx))m∈N\bigl(\int_{X}|Q_{m}x|^{2}\,\mu(dx)\bigr)_{m\in\mathbb{N}} has limit 00.

5. (Limit) The real sequence (W2(m)((pm)#μ,(pm)#ν))m∈N\bigl(W_{2}^{(m)}((p_{m})_{\#}\mu,(p_{m})_{\#}\nu)\bigr)_{m\in\mathbb{N}} has limit W2(μ,ν)W_{2}(\mu,\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…