Let $\gamma$ be the standard Gaussian measure on $\mathbb{ R}^n$ and let $\mathcal{ P}_{ \gamma}$ be the space of probability measures that are absolutely continuous with respect to $\gamma$. We study lower bounds for the functional $\mathcal{ F}_{ \gamma}(\mu) = { \rm Ent}(\mu) - \frac{ 1}{ 2} W^2_2(\mu, \nu)$, where $\mu \in \mathcal{ P}_{ \gamma}, \nu \in \mathcal{ P}_{ \gamma}$, ${ \rm Ent}(\mu) = \int \log\bigl( \frac{ \mu}{ \gamma}\bigr) d \mu$ is the relative Gaussian entropy, and $W_2$ is the quadratic Kantorovich distance. The minimizers of $\mathcal{ F}_{ \gamma}$ are solutions to a dimension-free Gaussian analog of the (real) K\"ahler-Einstein equation. We show that $\mathcal{ F}_{ \gamma}(\mu) $ is bounded from below under the assumption that the Gaussian Fisher information of $\nu$ is finite and prove a priori estimates for the minimizers. Our approach relies on certain stability estimates for the Gaussian log-Sobolev and Talagrand transportation inequalities.

Categories: math.FA

Moment measures and stability for Gaussian inequalities