TheoremBase

A Bound on the Gibbs Ornstein-Uhlenbeck Functional by Noise Gradients Gives Finite Fisher Information Relative to the Gibbs Measure

If the noise Ornstein-Uhlenbeck functional corrected by the pairing with the potential gradient is bounded by R times the noise-gradient norm on bounded C2C^2 cylindrical functions, the measure has a relative score with respect to the Gibbs measure with weighted Fisher information at most R squared.

Statement

In the setting of A Diagonal Gaussian Reference Measure on the Noise Wasserstein Space, Rescaled Heads and Gaussian Tails: Standing Notation, let VV be an admissible cylindrical potential with noise gradient ∇aV\nabla_{a}V, let β∈R\beta\in\mathbb{R} be positive, let γβV\gamma^{V}_{\beta} be the Gibbs measure of VV at temperature β\beta, let μ∈P2(X)\mu\in\mathcal{P}_{2}(X), the set of The Second Moment of a Borel Probability Measure on a Hilbert Space and the Probability Measures with Finite Second Moment §space, be such that VV is integrable with respect to μ\mu, and let R∈RR\in\mathbb{R} be nonnegative. By Basic Properties of an Admissible Cylindrical Potential: Continuity, Growth under Noise Translations, Integrability, the Tangent Inequality and Tangency of the Noise Gradient §integrable, ∣∇aV∣a|\nabla_{a}V|_{a} and each ∂kV\partial_{k}V are integrable with respect to μ\mu, so The Relative Score with Respect to the Gibbs Measure of an Admissible Cylindrical Potential and The Weighted Fisher Information Relative to the Gibbs Measure of an Admissible Cylindrical Potential apply to μ\mu. For n∈Nn\in\mathbb{N}, Cb2(Rn)C^{2}_{b}(\mathbb{R}^{n}) is the set of Bounded Twice Continuously Differentiable Functions with Bounded First and Second Partial Derivatives on Euclidean Space §bounded with q=nq=n; for g∈Cb2(Rn)g\in C^{2}_{b}(\mathbb{R}^{n}), Lμa,V(g)L^{a,V}_{\mu}(g) is the Gibbs Ornstein-Uhlenbeck functional of μ\mu at gg with potential VV and temperature β\beta, which applies since μ∈P2(X)\mu\in\mathcal{P}_{2}(X) and VV is integrable with respect to μ\mu; g∘png\circ p_{n} is a bounded C2C^{2} cylindrical function, and its noise gradient ∇a(g∘pn)\nabla_{a}(g\circ p_{n}) is bounded in ∣⋅∣a|\cdot|_{a} by that clause. ∥⋅∥μ\lVert\cdot\rVert_{\mu} is the norm of L2(μ;Xa)L^{2}(\mu;X^{a}), the space of Probability Measures on a Hilbert Space Transported in the Noise Norm: Standing Notation §fields, and the class of ∇a(g∘pn)\nabla_{a}(g\circ p_{n}) in it is again written ∇a(g∘pn)\nabla_{a}(g\circ p_{n}). Suppose that

∣Lμa,V(g)∣≤R ∥∇a(g∘pn)∥μfor every n∈N and every g∈Cb2(Rn).\bigl|L^{a,V}_{\mu}(g)\bigr|\le R\,\lVert\nabla_{a}(g\circ p_{n})\rVert_{\mu}\qquad\text{for every }n\in\mathbb{N}\text{ and every }g\in C^{2}_{b}(\mathbb{R}^{n}).

(Finite Fisher information relative to the Gibbs measure) Then μ\mu has a relative score with respect to γβV\gamma^{V}_{\beta} and finite Fisher information relative to γβV\gamma^{V}_{\beta} with weights aa, and Ia(μ ∣ γβV)≤R2\mathcal{I}_{a}(\mu\,|\,\gamma^{V}_{\beta})\le R^{2}.

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…