TheoremBase

A Log-Sobolev Inequality Survives a Bounded Perturbation of the Measure (Holley-Stroock)

Holley-Stroock perturbation: if a probability measure μ\mu satisfies a log-Sobolev inequality with constant CC for bounded C1C^1 cylindrical functions and noise gradients, then the normalized measure with density proportional to e−Ue^{-U}, for a bounded Borel UU with values in [u−,u+][u_-,u_+], satisfies it with constant Ceu+−u−Ce^{u_+-u_-}.

Statement

In the setting of Gaussian Analysis Relative to a Diagonal Gaussian Reference Measure with Noise Weights: Standing Notation, with 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, for which ∣∇aF∣a2|\nabla_{a}F|_{a}^{2} is a bounded nonnegative Borel function by The Noise Gradient of a Cylindrical Function and the Noise Tangent Space at a Probability Measure on a Hilbert Space §gradient; for ν∈P(X)\nu\in\mathcal{P}(X), Ent⁡ν\operatorname{Ent}_{\nu} is the entropy with respect to ν\nu, and ϕ\phi is 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. exp⁡\exp is the exponential function. For every F∈FCb1(X)F\in\mathcal{F}C^{1}_{b}(X) and every θ∈P(X)\theta\in\mathcal{P}(X), the functions F2F^{2} and ϕ∘F2\phi\circ F^{2} are bounded and Borel, hence integrable with respect to θ\theta, so that Ent⁡θ(F2)\operatorname{Ent}_{\theta}(F^{2}) is defined: FF is bounded and Borel by Bounded C^1 Cylindrical Functions: Linear Structure, the Gradient and the Partial Derivatives, Bounds and Integrability, and Density in the Square-Integrable Functions §bounded-borel, so F2F^{2} is Borel by claim 3 of Arithmetic, Absolute Values, and Pointwise Limits of Measurable Real-Valued Functions and its values lie in an interval [0,b][0,b]; ϕ\phi is continuous by The Function slog⁡ss\log s: Continuity, Young's Inequality and Lower Bounds, with the Elementary Bounds for the Exponential and the Logarithm §continuous, so ϕ∘F2\phi\circ F^{2} is Borel by that clause and bounded on [0,b][0,b] by Extreme Value Theorem on a Closed Real Interval; and bounded Borel functions are integrable with respect to a probability measure by claim 6 of Borel Measurability and Bounded Integration on a Metric Space.

Let μ∈P(X)\mu\in\mathcal{P}(X) and let C∈RC\in\mathbb{R} with 0≤C0\le C be such that

Ent⁡μ(F2)≤2C∫X∣∇aF∣a2 dμfor every F∈FCb1(X).\operatorname{Ent}_{\mu}(F^{2})\le2C\int_{X}|\nabla_{a}F|_{a}^{2}\,d\mu\qquad\text{for every }F\in\mathcal{F}C^{1}_{b}(X).

Let U:X→RU:X\to\mathbb{R} be Borel and bounded, and let u−,u+∈Ru_{-},u_{+}\in\mathbb{R} satisfy u−≤U(x)≤u+u_{-}\le U(x)\le u_{+} for every x∈Xx\in X.

1. (The perturbed measure) The function exp⁡(−U)\exp(-U) is Borel and bounded, hence integrable with respect to μ\mu; the number Z=∫Xexp⁡(−U) dμZ=\int_{X}\exp(-U)\,d\mu satisfies exp⁡(−u+)≤Z≤exp⁡(−u−)\exp(-u_{+})\le Z\le\exp(-u_{-}), so Z>0Z>0; the function ν\nu on B(X)\mathcal{B}(X) given by ν(A)=Z−1∫X1Aexp⁡(−U) dμ\nu(A)=Z^{-1}\int_{X}\mathbf{1}_{A}\exp(-U)\,d\mu belongs to P(X)\mathcal{P}(X); and for every bounded Borel f:X→Rf:X\to\mathbb{R}, ∫Xf dν=Z−1∫Xfexp⁡(−U) dμ\int_{X}f\,d\nu=Z^{-1}\int_{X}f\exp(-U)\,d\mu.

2. (Perturbed log-Sobolev inequality) For every F∈FCb1(X)F\in\mathcal{F}C^{1}_{b}(X),

Ent⁡ν(F2)≤2Cexp⁡(u+−u−)∫X∣∇aF∣a2 dν.\operatorname{Ent}_{\nu}(F^{2})\le2C\exp(u_{+}-u_{-})\int_{X}|\nabla_{a}F|_{a}^{2}\,d\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…