TheoremBase

Talagrand, HWI and Log-Sobolev Inequalities for a Uniformly Displacement Convex Noise Penalty Pair with a Point of Zero Score

For a λ\lambda-displacement convex noise penalty pair with λ>0\lambda>0 and a measure μ∗\mu_* at which the score vanishes, the Talagrand inequality E(μ∗)+λ2Wa(μ∗,ν)2≤E(ν)\mathcal{E}(\mu_*)+\frac{\lambda}{2}W_a(\mu_*,\nu)^2\le\mathcal{E}(\nu), the HWI inequality, and the log-Sobolev inequality E(μ)−E(μ∗)≤12λ∥Σ(μ)∥μ2\mathcal{E}(\mu)-\mathcal{E}(\mu_*)\le\frac{1}{2\lambda}\lVert\Sigma(\mu)\rVert_\mu^2 hold.

Statement

In the setting of First-Order Equations on the Noise Wasserstein Space Relative to a Noise Penalty Pair: Standing Notation, let (D,DΣ,E,Σ)(\mathcal{D},\mathcal{D}_{\Sigma},\mathcal{E},\Sigma) be a noise penalty pair on Pρa\mathcal{P}^{a}_{\rho}, let λ∈R\lambda\in\mathbb{R} be positive, and suppose that the pair is λ\lambda-displacement convex. Let μ∗∈DΣ\mu_{*}\in\mathcal{D}_{\Sigma} be such that Σ(μ∗)\Sigma(\mu_{*}) is the zero vector of L2(μ∗;Xa)L^{2}(\mu_{*};X^{a}). For μ∈DΣ\mu\in\mathcal{D}_{\Sigma}, Σ(μ)∈L2(μ;Xa)\Sigma(\mu)\in L^{2}(\mu;X^{a}) by Noise Penalty Pairs on the Noise Wasserstein Space: the Penalty, Its Score, and Their Domains §pair, with norm ∥Σ(μ)∥μ\lVert\Sigma(\mu)\rVert_{\mu}; and WaW_{a} is the noise Wasserstein distance of First-Order Equations on the Noise Wasserstein Space Relative to a Noise Penalty Pair: Standing Notation §space, defined on D⊆Pρa\mathcal{D}\subseteq\mathcal{P}^{a}_{\rho}.

1. (Talagrand inequality) For every ν∈D\nu\in\mathcal{D},

E(μ∗)+λ2 Wa(μ∗,ν)2≤E(ν);\mathcal{E}(\mu_{*})+\frac{\lambda}{2}\,W_{a}(\mu_{*},\nu)^{2}\le\mathcal{E}(\nu);

in particular E(μ∗)≤E(ν)\mathcal{E}(\mu_{*})\le\mathcal{E}(\nu) for every ν∈D\nu\in\mathcal{D}.

2. (HWI inequality) For every μ∈DΣ\mu\in\mathcal{D}_{\Sigma},

E(μ)−E(μ∗)≤∥Σ(μ)∥μ Wa(μ,μ∗)−λ2 Wa(μ,μ∗)2.\mathcal{E}(\mu)-\mathcal{E}(\mu_{*})\le\lVert\Sigma(\mu)\rVert_{\mu}\,W_{a}(\mu,\mu_{*})-\frac{\lambda}{2}\,W_{a}(\mu,\mu_{*})^{2}.

3. (Log-Sobolev inequality) For every μ∈DΣ\mu\in\mathcal{D}_{\Sigma},

E(μ)−E(μ∗)≤12λ ∥Σ(μ)∥μ2.\mathcal{E}(\mu)-\mathcal{E}(\mu_{*})\le\frac{1}{2\lambda}\,\lVert\Sigma(\mu)\rVert_{\mu}^{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…