TheoremBase

The Gaussian Sobolev Space: Linear Structure, Closedness, Hermite Polynomials and the Number-Operator Form of the Dirichlet Form, and the Log-Sobolev and Poincare Inequalities

The Gaussian Sobolev space is a linear subspace of L2(γc)L^2(\gamma_c) on which the noise gradient is linear and closed; it contains the cylindrical Hermite polynomials, with explicit gradient coordinates, and on them the Dirichlet form is diagonal with Ea(Hα,Hα)=(θ⋅α)α!cα\mathcal{E}_a(H_\alpha,H_\alpha)=(\theta\cdot\alpha)\alpha!c^\alpha; and if ck≤κakc_k\le\kappa a_k for all kk, every element satisfies the log-Sobolev inequality with constant 2κ2\kappa and the Poincare inequality with constant κ\kappa.

Statement

In the setting of Gaussian Analysis Relative to a Diagonal Gaussian Reference Measure with Noise Weights: Standing Notation, with the Gaussian Sobolev space Da1,2\mathbb{D}^{1,2}_{a}, the noise gradient ∇aF\nabla_{a}F of F∈Da1,2F\in\mathbb{D}^{1,2}_{a} of The Gaussian Sobolev Space of the Noise Gradient, and the Dirichlet form Ea\mathcal{E}_{a}; the coordinate of an element of L2(γc;Xa)L^{2}(\gamma_{c};X^{a}) along fkf_{k}, an element of L2(γc)L^{2}(\gamma_{c}), is that of The Space of Square-Integrable Hilbert-Valued Maps is a Real Hilbert Space: Coordinates and Synthesis §coordinates; the rates θk\theta_{k} of Gaussian Analysis Relative to a Diagonal Gaussian Reference Measure with Noise Weights: Standing Notation §rates and θ=(θk)k∈N\theta=(\theta_{k})_{k\in\mathbb{N}}; the multi-indices α∈A\alpha\in\mathcal{A} with α!\alpha!, cαc^{\alpha}, θ⋅α\theta\cdot\alpha, εk\varepsilon_{k} and α−εk\alpha-\varepsilon_{k}, and the cylindrical Hermite polynomials HαH_{\alpha}; and, γc\gamma_{c} being a probability measure, the entropy Ent⁡γc\operatorname{Ent}_{\gamma_{c}}, with the function ϕ\phi of The Function slog⁡ss\log s: Continuity, Young's Inequality and Lower Bounds, with the Elementary Bounds for the Exponential and the Logarithm.

1. (Linear structure) Da1,2\mathbb{D}^{1,2}_{a} is a linear subspace of L2(γc)L^{2}(\gamma_{c}), and F↦∇aFF\mapsto\nabla_{a}F is linear on it.

2. (Closedness) Let (Fj)j∈N(F_{j})_{j\in\mathbb{N}} be a sequence in Da1,2\mathbb{D}^{1,2}_{a}, F∈L2(γc)F\in L^{2}(\gamma_{c}) and G∈L2(γc;Xa)G\in L^{2}(\gamma_{c};X^{a}) with ∥Fj−F∥2→0\lVert F_{j}-F\rVert_{2}\to0 and ∥∇aFj−G∥γc→0\lVert\nabla_{a}F_{j}-G\rVert_{\gamma_{c}}\to0. Then F∈Da1,2F\in\mathbb{D}^{1,2}_{a} and ∇aF=G\nabla_{a}F=G.

3. (Hermite polynomials) For every α∈A\alpha\in\mathcal{A}, the class of HαH_{\alpha}, which lies in L2(γc)L^{2}(\gamma_{c}) by Cylindrical Hermite Polynomials: Growth and Integrability, Partial Derivatives, Orthogonality, and Monomials as Hermite Combinations §cylindrical and is again written HαH_{\alpha}, belongs to Da1,2\mathbb{D}^{1,2}_{a}, and for every k∈Nk\in\mathbb{N} the coordinate of ∇aHα\nabla_{a}H_{\alpha} along fkf_{k} is ak1/2αkHα−εka_{k}^{1/2}\alpha_{k}H_{\alpha-\varepsilon_{k}} if αk≥1\alpha_{k}\ge1 and 00 if αk=0\alpha_{k}=0.

4. (Number-operator form) For all α,β∈A\alpha,\beta\in\mathcal{A},

Ea(Hα,Hβ)={(θ⋅α) α! cαif α=β,0if α≠β.\mathcal{E}_{a}(H_{\alpha},H_{\beta})=\begin{cases}(\theta\cdot\alpha)\,\alpha!\,c^{\alpha}&\text{if }\alpha=\beta,\\0&\text{if }\alpha\ne\beta.\end{cases}

5. (Log-Sobolev inequality) Let κ∈R\kappa\in\mathbb{R} be positive with ck≤κ akc_{k}\le\kappa\,a_{k} for every k∈Nk\in\mathbb{N}, and let F∈Da1,2F\in\mathbb{D}^{1,2}_{a} with a Borel representative ff. Then f2f^{2} and ϕ∘f2\phi\circ f^{2} are integrable with respect to γc\gamma_{c}, and

Ent⁡γc(f2)≤2κ Ea(F,F).\operatorname{Ent}_{\gamma_{c}}(f^{2})\le2\kappa\,\mathcal{E}_{a}(F,F).

6. (Poincar'e inequality) For κ\kappa, FF and ff as in claim 5, ff is integrable with respect to γc\gamma_{c}, and

∫Xf2 dγc−(∫Xf dγc)2≤κ Ea(F,F).\int_{X}f^{2}\,d\gamma_{c}-\Bigl(\int_{X}f\,d\gamma_{c}\Bigr)^{2}\le\kappa\,\mathcal{E}_{a}(F,F).

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…