Transportation inequalities for Markov kernels and their applications (Q2042801)

From MaRDI portal





scientific article
Language Label Description Also known as
default for all languages
No label defined
    English
    Transportation inequalities for Markov kernels and their applications
    scientific article

      Statements

      Transportation inequalities for Markov kernels and their applications (English)
      0 references
      0 references
      0 references
      21 July 2021
      0 references
      The aim of the paper is to study the connection between reverse Poincaré inequalities and Hellinger-Kantorovich contraction inequalities for heat semigroups associated with Markovian kernels. One of the main results of the paper (Theorem 3.7) is that given a heat kernel \(P\) on a metric space \(X\) (assumed to be a length space, in particular, path connected, equipped with a strong upper gradient) then the following inequalities are equivalent: the reverse Poincaré inequality \[ | \nabla P \, f|^2 \le C (P(f^2)-(P(f))^2), \quad \forall \, f \in \mathrm{Lip}_{b}(X), \] the Hellinger-Kantorovich contraction inequalities \[ He_{2}(\mu_0 P, \mu_1 P) \le HK_{4/C}(\mu_0,\mu_1) \le \sqrt{\frac{C}{4}} W_{2}(\mu_0,\mu_1), \] where \(\mu_0,\mu_1\) are probability measures on \(X\) and \(He\), \(HK\), \(W_{2}\) are the Hellinger, the Hellinger-Kantorovich and the Wasserstein distances respectively, and the Harnack type inequality \[ Pf(x) \le P f(y) + \sqrt{C} \, d(x,y) \sqrt{P(f^2)(x)}, \quad\forall \, x,y \in X, \, f \in B_{b}(X), \quad f \ge 0. \] This result builds on the following tools: \begin{itemize} \item[(1)] A suitable extension of recent results of \textit{G. Luise} and \textit{G. Savaré} [Discrete Contin. Dyn. Syst., Ser. S 14, No. 1, 273--297 (2021; Zbl 1459.49030)] on the contraction properties of heat semigroups in spaces of measures, to general metric spaces; \item[(2)] The connection of the resulting Hellinger-Kantorovich contraction inequalities to reverse Poincaré inequalities; \item[(3)] A dynamic dual formulation of the Hellinger-Kantorovich distance to define a new family of divergences on \(P(X)\) which generalize the Rényi divergence. This allows the authors to prove that contraction inequalities for these divergences are equivalent to the reverse logarithmic Sobolev and Wang-Harnack inequalities. \end{itemize} This important result connects the Hellinger-Kantorovich contraction with well known functional inequalities. The results of the paper are illustrated by various important applications, including subelliptic diffusions arising in sub-Riemannian geometry, non-symmetric Ornstein-Uhlenbeck operators on Carnot groups, Langevin dynamics driven by Lévy processes, and others.
      0 references
      functional inequalities
      0 references
      Hellinger distance
      0 references
      Kantorovich-Wasserstein distance
      0 references
      Kuwada duality
      0 references
      Markov kernels
      0 references
      optimal transport
      0 references
      reverse logarithmic Sobolev inequality
      0 references
      reverse Poincaré inequality
      0 references
      0 references
      0 references
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references