TheoremBase

Diagonal Quadratic Profiles on the Noise Wasserstein Space: Moment Bounds, a Lipschitz Estimate, the Gradient Field, and Noise Intrinsic Test Functions

Integrals of diagonal quadratics with suitably summable coefficients are finite on all measures noise-connected to a diagonal Gaussian, depend Lipschitz-continuously on the measure, and are noise intrinsic test functions whose gradient is the corresponding linear field.

Statement

In the setting of A Diagonal Gaussian Reference Measure on the Noise Wasserstein Space, Rescaled Heads and Gaussian Tails: Standing Notation, so that the reference measure is ρ=γc\rho=\gamma_{c}, every μ∈Pρa\mu\in\mathcal{P}^{a}_{\rho} belongs to the set P2(X)\mathcal{P}_{2}(X) of The Second Moment of a Borel Probability Measure on a Hilbert Space and the Probability Measures with Finite Second Moment §space 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 §moments, so that for every k∈Nk\in\mathbb{N} the Borel function x↦xk2x\mapsto x_{k}^{2}, which is at most ∣x∣2|x|^{2}, is integrable with respect to μ\mu. For μ,ν∈Pρa\mu,\nu\in\mathcal{P}^{a}_{\rho}, Πa(μ,ν)\Pi^{a}(\mu,\nu) is the set of couplings of finite noise cost and Ia(π)I^{a}(\pi) the noise cost of Couplings of Finite Noise Cost and Their Noise Cost §cost, and Ja(η,π)\mathcal{J}^{a}(\eta,\pi) is the noise displacement pairing; the coordinate of an element of L2(μ;Xa)L^{2}(\mu;X^{a}) along fkf_{k} is that of The Space of Square-Integrable Hilbert-Valued Maps is a Real Hilbert Space: Coordinates and Synthesis §coordinates, and ⟨⋅,⋅⟩μ\langle\cdot,\cdot\rangle_{\mu}, ∥⋅∥μ\lVert\cdot\rVert_{\mu} are the inner product and norm of L2(μ;Xa)L^{2}(\mu;X^{a}). Noise intrinsic test functions and gradients along noise couplings are those of those definitions, square roots are those of Existence and Uniqueness of the Nonnegative Square Root, and convergent series are those of that definition. A real sequence b=(bk)k∈Nb=(b_{k})_{k\in\mathbb{N}} is called admissible if the series ∑k=1∞∣bk∣ ck\sum_{k=1}^{\infty}|b_{k}|\,c_{k} converges and there is B∈RB\in\mathbb{R} with 0≤B0\le B and ∣bk∣ ak≤B|b_{k}|\,a_{k}\le B for every k∈Nk\in\mathbb{N}; such a BB is called a bound for bb.

1. (Moments) Let t=(tk)k∈Nt=(t_{k})_{k\in\mathbb{N}} be a sequence of nonnegative real numbers such that ∑k=1∞tkck\sum_{k=1}^{\infty}t_{k}c_{k} converges, and let M∈RM\in\mathbb{R} satisfy 0≤M0\le M and tkak≤Mt_{k}a_{k}\le M for every k∈Nk\in\mathbb{N}. For every μ∈Pρa\mu\in\mathcal{P}^{a}_{\rho} the series St(μ)=∑k=1∞tk∫Xxk2 μ(dx)S_{t}(\mu)=\sum_{k=1}^{\infty}t_{k}\int_{X}x_{k}^{2}\,\mu(dx) converges, and

St(μ)≤2∑k=1∞tkck+2M Wa(μ,ρ)2.S_{t}(\mu)\le2\sum_{k=1}^{\infty}t_{k}c_{k}+2M\,W_{a}(\mu,\rho)^{2}.

2. (A Lipschitz estimate) Let tt and MM be as in claim 1, and let μ,ν∈Pρa\mu,\nu\in\mathcal{P}^{a}_{\rho} and π∈Πa(μ,ν)\pi\in\Pi^{a}(\mu,\nu). Then the series ∑k=1∞tk∫X×X(yk−xk)2 π(d(x,y))\sum_{k=1}^{\infty}t_{k}\int_{X\times X}(y_{k}-x_{k})^{2}\,\pi(d(x,y)) converges with sum at most M Ia(π)M\,I^{a}(\pi), and

∣St(μ)−St(ν)∣≤M Ia(π).\Bigl|\sqrt{S_{t}(\mu)}-\sqrt{S_{t}(\nu)}\Bigr|\le\sqrt{M}\,\sqrt{I^{a}(\pi)} .

3. (The gradient field) Let bb be admissible and μ∈Pρa\mu\in\mathcal{P}^{a}_{\rho}. There is exactly one Vb(μ)∈L2(μ;Xa)V_{b}(\mu)\in L^{2}(\mu;X^{a}) whose coordinate along fkf_{k} is, for every k∈Nk\in\mathbb{N}, the class of the function x↦ak1/2 bk xkx\mapsto a_{k}^{1/2}\,b_{k}\,x_{k}, and

∥Vb(μ)∥μ2=∑k=1∞ak bk2∫Xxk2 μ(dx).\lVert V_{b}(\mu)\rVert_{\mu}^{2}=\sum_{k=1}^{\infty}a_{k}\,b_{k}^{2}\int_{X}x_{k}^{2}\,\mu(dx).

4. (The profile) Let bb be admissible. For every μ∈Pρa\mu\in\mathcal{P}^{a}_{\rho} the series ∑k=1∞bk∫Xxk2 μ(dx)\sum_{k=1}^{\infty}b_{k}\int_{X}x_{k}^{2}\,\mu(dx) converges absolutely, and the diagonal quadratic profile with coefficients bb is the function

Φb:Pρa→R,Φb(μ)=12∑k=1∞bk∫Xxk2 μ(dx).\Phi_{b}:\mathcal{P}^{a}_{\rho}\to\mathbb{R},\qquad\Phi_{b}(\mu)=\frac{1}{2}\sum_{k=1}^{\infty}b_{k}\int_{X}x_{k}^{2}\,\mu(dx).

5. (First-order expansion) Let bb be admissible with bound BB. For μ,ν∈Pρa\mu,\nu\in\mathcal{P}^{a}_{\rho} and π∈Πa(μ,ν)\pi\in\Pi^{a}(\mu,\nu),

∣Φb(ν)−Φb(μ)−Ja(Vb(μ),π)∣≤B2 Ia(π).\bigl|\Phi_{b}(\nu)-\Phi_{b}(\mu)-\mathcal{J}^{a}(V_{b}(\mu),\pi)\bigr|\le\frac{B}{2}\,I^{a}(\pi).

6. (Test function) Let bb be admissible. For every subset Q⊆PρaQ\subseteq\mathcal{P}^{a}_{\rho}, Φb\Phi_{b} is a noise intrinsic test function on QQ, and ∇Φb(μ)=Vb(μ)\nabla\Phi_{b}(\mu)=V_{b}(\mu) for every μ∈Q\mu\in Q.

7. (Scaling) Let bb be admissible and s∈Rs\in\mathbb{R}. Then the sequence sb=(s bk)k∈Nsb=(s\,b_{k})_{k\in\mathbb{N}} is admissible, and Φsb(μ)=s Φb(μ)\Phi_{sb}(\mu)=s\,\Phi_{b}(\mu) and Vsb(μ)=s Vb(μ)V_{sb}(\mu)=s\,V_{b}(\mu) for every μ∈Pρa\mu\in\mathcal{P}^{a}_{\rho}.

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…