TheoremBase

Noise Intrinsic Test Functions Pulled Back along a Mode Restriction, and the Squared Coarse Noise Wasserstein Distance as a Fine Test Function

Composing a coarse-level noise intrinsic test function with push-forward by the mode restriction gives a fine-level test function whose gradient is the pulled-back coarse gradient; in particular the squared coarse distance from the restricted measure to a fixed measure is a fine test function with an explicit gradient of norm twice that distance.

Statement

In the setting of The Real Numbers: Standing Notation and Background and Two Noise Levels over Hilbert Spaces: a Fine and a Coarse Reading of the Noise Framework and Level-Indexed Notation, let levels 1 and 2 be as in Two Noise Levels over Hilbert Spaces: a Fine and a Coarse Reading of the Noise Framework and Level-Indexed Notation §levels, with the level-indexed notation of Two Noise Levels over Hilbert Spaces: a Fine and a Coarse Reading of the Noise Framework and Level-Indexed Notation §notation, and let (κ,p,j)(\kappa,p,j) be a mode-restriction link from level 1 to level 2, with mode restriction pp and mode embedding jj. Suppose that p#ρ1∈Pρ2ap_{\#}\rho_{1}\in\mathcal{P}^{a}_{\rho_{2}}, so that p#μ∈Pρ2ap_{\#}\mu\in\mathcal{P}^{a}_{\rho_{2}} for every μ∈Pρ1a\mu\in\mathcal{P}^{a}_{\rho_{1}} by Properties of a Mode-Restriction Link: Isometric Embedding, Contraction of Noise Norms and Noise Wasserstein Distances, and Pull-Back of Cylindrical Functions and Tangent Fields §space; for φ:Pρ2a→R\varphi:\mathcal{P}^{a}_{\rho_{2}}\to\mathbb{R}, φ∘p#:Pρ1a→R\varphi\circ p_{\#}:\mathcal{P}^{a}_{\rho_{1}}\to\mathbb{R} is the function μ↦φ(p#μ)\mu\mapsto\varphi(p_{\#}\mu). Let Q1⊆Pρ1a\mathcal{Q}_{1}\subseteq\mathcal{P}^{a}_{\rho_{1}} and Q2⊆Pρ2a\mathcal{Q}_{2}\subseteq\mathcal{P}^{a}_{\rho_{2}} satisfy p#μ∈Q2p_{\#}\mu\in\mathcal{Q}_{2} for every μ∈Q1\mu\in\mathcal{Q}_{1}. Noise intrinsic test functions on Qi\mathcal{Q}_{i} and their gradients along noise couplings, written ∇i\nabla_{i}, are taken at level ii; for μ∈P(X1)\mu\in\mathcal{P}(X_{1}) and η∈L2(p#μ;X2a)\eta\in L^{2}(p_{\#}\mu;X_{2}^{a}), j∘η∘p∈L2(μ;X1a)j\circ\eta\circ p\in L^{2}(\mu;X_{1}^{a}) is the class of Properties of a Mode-Restriction Link: Isometric Embedding, Contraction of Noise Norms and Noise Wasserstein Distances, and Pull-Back of Cylindrical Functions and Tangent Fields §fields. The noise map property, noise-optimal maps SS and their displacements S−id2∈L2(σ;X2a)S-\mathrm{id}_{2}\in L^{2}(\sigma;X_{2}^{a}) for σ∈Pρ2a\sigma\in\mathcal{P}^{a}_{\rho_{2}} are taken at level 2, and id2−S\mathrm{id}_{2}-S denotes −(S−id2)-(S-\mathrm{id}_{2}). Then the following hold.

1. (Pull-back of test functions) Let φ\varphi be a noise intrinsic test function on Q2\mathcal{Q}_{2} at level 2. Then φ∘p#\varphi\circ p_{\#} is a noise intrinsic test function on Q1\mathcal{Q}_{1} at level 1, and

∇1(φ∘p#)(μ)=j∘∇2φ(p#μ)∘pfor every μ∈Q1.\nabla_{1}(\varphi\circ p_{\#})(\mu)=j\circ\nabla_{2}\varphi(p_{\#}\mu)\circ p\qquad\text{for every }\mu\in\mathcal{Q}_{1}.

2. (The squared coarse distance) Suppose that Q2\mathcal{Q}_{2} has the noise map property at level 2, and let ν0∈Pρ2a\nu_{0}\in\mathcal{P}^{a}_{\rho_{2}}. Let ψ:Pρ1a→R\psi:\mathcal{P}^{a}_{\rho_{1}}\to\mathbb{R} be the function

ψ(μ)=Wa,2(p#μ,ν0)2,\psi(\mu)=W_{a,2}(p_{\#}\mu,\nu_{0})^{2},

which is defined because p#μ,ν0∈Pρ2ap_{\#}\mu,\nu_{0}\in\mathcal{P}^{a}_{\rho_{2}} and any two members of Pρ2a\mathcal{P}^{a}_{\rho_{2}} are noise-connected at level 2 by The Noise Wasserstein Distance is a Metric on the Measures Noise-Connected to the Reference Measure: Existence of Noise-Optimal Couplings, Comparison with the Quadratic Wasserstein Distance and Lower Semicontinuity §connected. Then ψ\psi is a noise intrinsic test function on Q1\mathcal{Q}_{1} at level 1. For μ∈Q1\mu\in\mathcal{Q}_{1} a noise-optimal map SS from p#μp_{\#}\mu to ν0\nu_{0} at level 2 exists by the noise map property of Q2\mathcal{Q}_{2}, since p#μ∈Q2p_{\#}\mu\in\mathcal{Q}_{2}, and the class id2−S\mathrm{id}_{2}-S does not depend on its choice by The Noise Map Property of a Set of Probability Measures §map-property, taken at level 2. For every such μ\mu and SS,

∇1ψ(μ)=2 j∘(id2−S)∘p.\nabla_{1}\psi(\mu)=2\,j\circ(\mathrm{id}_{2}-S)\circ p.

For every μ∈Q1\mu\in\mathcal{Q}_{1},

∥∇1ψ(μ)∥μ,1=2 Wa,2(p#μ,ν0).\lVert\nabla_{1}\psi(\mu)\rVert_{\mu,1}=2\,W_{a,2}(p_{\#}\mu,\nu_{0}).

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…