TheoremBase

Log-Sobolev and Talagrand Inequalities for the Gibbs Measure of a Semiconvex Cylindrical Potential below the Critical Curvature (Bakry-Emery)

If ck≤κakc_k\le\kappa a_k for all kk and VV is an admissible cylindrical potential with semiconvexity constant KK where κK<β\kappa K<\beta, then the Gibbs measure of VV at temperature β\beta satisfies log-Sobolev inequalities in entropy form and in Gross's form and a Talagrand inequality in the noise Wasserstein distance, each with constant governed by κβ/(β−κK)\kappa\beta/(\beta-\kappa K).

Statement

In the setting of Gaussian Analysis Relative to a Diagonal Gaussian Reference Measure with Noise Weights: Standing Notation, with the noise Wasserstein distance WaW_{a} on Pρa\mathcal{P}^{a}_{\rho} and relative entropy H(⋅ ∣ ⋅)H(\cdot\,|\,\cdot) of Gaussian Analysis Relative to a Diagonal Gaussian Reference Measure with Noise Weights: Standing Notation §background, and FCb1(X)\mathcal{F}C^{1}_{b}(X) and the noise gradients ∇aF\nabla_{a}F of Gaussian Analysis Relative to a Diagonal Gaussian Reference Measure with Noise Weights: Standing Notation §cylindrical. Let κ,β∈R\kappa,\beta\in\mathbb{R} be positive with ck≤κ akc_{k}\le\kappa\,a_{k} for every k∈Nk\in\mathbb{N}, let K∈RK\in\mathbb{R} satisfy 0≤K0\le K and κK<β\kappa K<\beta, and let VV be an admissible cylindrical potential with semiconvexity constant KK. γβV\gamma^{V}_{\beta} is the Gibbs measure of VV at temperature β\beta; the relative score with respect to γβV\gamma^{V}_{\beta} and finite Fisher information relative to γβV\gamma^{V}_{\beta} with weights aa, with value Ia(⋅ ∣ γβV)\mathcal{I}_{a}(\cdot\,|\,\gamma^{V}_{\beta}), are those of the cited definitions, applicable to every μ\mu of finite relative entropy with respect to γβV\gamma^{V}_{\beta} by Entropy and Score Relative to the Gibbs Measure Split into Their Gaussian Parts and the Potential, with a Fisher Information Bound §domain; and Ent⁡γβV\operatorname{Ent}_{\gamma^{V}_{\beta}} is the entropy with respect to γβV\gamma^{V}_{\beta}.

1. (Log-Sobolev inequality, entropy form) Let μ∈P(X)\mu\in\mathcal{P}(X) have finite relative entropy with respect to γβV\gamma^{V}_{\beta}, a relative score with respect to γβV\gamma^{V}_{\beta}, and finite Fisher information relative to γβV\gamma^{V}_{\beta} with weights aa. Then

H(μ ∣ γβV)≤κβ2(β−κK) Ia(μ ∣ γβV).H(\mu\,|\,\gamma^{V}_{\beta})\le\frac{\kappa\beta}{2(\beta-\kappa K)}\,\mathcal{I}_{a}(\mu\,|\,\gamma^{V}_{\beta}).

2. (Talagrand inequality) Let μ∈P(X)\mu\in\mathcal{P}(X) have finite relative entropy with respect to γβV\gamma^{V}_{\beta}. Then μ\mu and γβV\gamma^{V}_{\beta} belong to Pρa\mathcal{P}^{a}_{\rho}, and

Wa(μ,γβV)2≤2κββ−κK H(μ ∣ γβV).W_{a}(\mu,\gamma^{V}_{\beta})^{2}\le\frac{2\kappa\beta}{\beta-\kappa K}\,H(\mu\,|\,\gamma^{V}_{\beta}).

3. (Log-Sobolev inequality, Gross's form) For every F∈FCb1(X)F\in\mathcal{F}C^{1}_{b}(X), the functions F2F^{2} and ϕ∘F2\phi\circ F^{2}, with ϕ\phi the function of The Function slog⁡ss\log s: Continuity, Young's Inequality and Lower Bounds, with the Elementary Bounds for the Exponential and the Logarithm, are bounded and Borel, hence integrable with respect to γβV\gamma^{V}_{\beta}, and

Ent⁡γβV(F2)≤2κββ−κK∫X∣∇aF∣a2 dγβV.\operatorname{Ent}_{\gamma^{V}_{\beta}}(F^{2})\le\frac{2\kappa\beta}{\beta-\kappa K}\int_{X}|\nabla_{a}F|_{a}^{2}\,d\gamma^{V}_{\beta}.

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…