TheoremBase

Well-Posedness of the Hamilton-Jacobi Equation with Gibbs Score Drift for a Semiconvex Cylindrical Potential: Comparison, Existence and Uniqueness of a Bounded Viscosity Solution

For every admissible cylindrical potential, however nonconvex, every temperature, discount, control cost in (0,1] and bounded uniformly continuous running cost, the Hamilton-Jacobi equation with Gibbs score drift satisfies comparison and has exactly one bounded viscosity solution.

Statement

In the settings of The Real Numbers: Standing Notation and Background and A Diagonal Gaussian Reference Measure on the Noise Wasserstein Space, Rescaled Heads and Gaussian Tails: Standing Notation, so that the reference measure is ρ=γc\rho=\gamma_{c}, let K∈RK\in\mathbb{R} be nonnegative and let VV be an admissible cylindrical potential with semiconvexity constant KK; no relation between KK and β/κ\beta/\kappa is assumed; let β\beta and κ\kappa be positive real numbers with ck≤κ akc_{k}\le\kappa\,a_{k} for every k∈Nk\in\mathbb{N}, and let (D,DΣ,E,Σ)(\mathcal{D},\mathcal{D}_{\Sigma},\mathcal{E},\Sigma) be the Gibbs entropy pair with potential VV and temperature β\beta, whose hypothesis holds with this κ\kappa; it is a noise penalty pair on Pρa\mathcal{P}^{a}_{\rho} by The Gibbs Entropy Pair is a Noise Penalty Pair, with Nonnegative Penalty and Dense Score Domain §pair, and its penalty domain D\mathcal{D} is the set of the measures of finite relative entropy with respect to the Gibbs measure γβV\gamma^{V}_{\beta} of VV at temperature β\beta by The Gibbs Entropy Pair on the Noise Wasserstein Space: Relative Entropy to the Gibbs Measure and the Gibbs Score Field §penalty-domain. WaW_{a} is the noise Wasserstein distance and ∣s∣|s| the absolute value of s∈Rs\in\mathbb{R}.

(The data) λ0,θ∈R\lambda_{0},\theta\in\mathbb{R} satisfy 0<λ00<\lambda_{0} and 0<θ≤10<\theta\le1, and g:D→Rg:\mathcal{D}\to\mathbb{R} is bounded and uniformly continuous on D\mathcal{D}, relative to D\mathcal{D} in the metric space (Pρa,Wa)(\mathcal{P}^{a}_{\rho},W_{a}) of The Noise Wasserstein Distance is a Metric on the Measures Noise-Connected to the Reference Measure: Existence of Noise-Optimal Couplings, Comparison with the Quadratic Wasserstein Distance and Lower Semicontinuity §metric and with R\mathbb{R} carrying the metric of The Absolute Value Metric on the Real Line. Fix C∈RC\in\mathbb{R} with 0≤C0\le C and ∣g(μ)∣≤C|g(\mu)|\le C for every μ∈D\mu\in\mathcal{D}; such a CC exists, since if C′C' bounds ∣g∣|g| on D\mathcal{D} as in Bounded Real-Valued Function on a Set, the larger of C′C' and 00 serves.

Viscosity subsolutions, supersolutions and solutions are those of the Hamilton-Jacobi equation with Gibbs score drift relative to γβV\gamma^{V}_{\beta}, with discount λ0\lambda_{0}, control cost θ\theta and running cost gg.

1. (Comparison) Let u,w:D→Ru,w:\mathcal{D}\to\mathbb{R} and b,b′∈Rb,b'\in\mathbb{R} satisfy u(μ)≤bu(\mu)\le b and b′≤w(μ)b'\le w(\mu) for every μ∈D\mu\in\mathcal{D}; the pair being noise-closed by The Gibbs Entropy Pair is Noise-Closed, and Its Penalty Bounds the Squared Noise Wasserstein Distance to the Reference Measure §noise-closed, Basic Properties of the Delta-Envelopes on the Noise Wasserstein Space, and the Envelopes of Bounded Functions for a Noise-Closed Penalty Pair §growth shows that uu has penalty-subordinate growth from above and ww from below. Let uu be a viscosity subsolution and ww a viscosity supersolution. Then u(μ)≤w(μ)u(\mu)\le w(\mu) for every μ∈D\mu\in\mathcal{D}.

2. (Existence) There is a viscosity solution u:D→Ru:\mathcal{D}\to\mathbb{R} with −λ0−1C≤u(μ)≤λ0−1C-\lambda_{0}^{-1}C\le u(\mu)\le\lambda_{0}^{-1}C for every μ∈D\mu\in\mathcal{D}, λ0−1\lambda_{0}^{-1} being the multiplicative inverse of λ0\lambda_{0}.

3. (Uniqueness) Let u,u′:D→Ru,u':\mathcal{D}\to\mathbb{R} be viscosity solutions, each bounded. Then u(μ)=u′(μ)u(\mu)=u'(\mu) for every μ∈D\mu\in\mathcal{D}; together with claim 2, the equation has exactly one bounded viscosity solution.

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…