A characterization of dimension free concentration in terms of transportation inequalities (Q971948): Difference between revisions

A probability measure \(\mu\) on \(\mathbb{R}^{k}\) has the Gaussian dimension free concentration property if there are \(a \geqslant 0\) and \(M \geqslant 0\) such that for any \(n \in \mathbb{N}\) the product measure \(\mu^{n}\) satisfies \[ \mu^{n}\left( A+rB\right)\geqslant 1-Me^{-ar^{2}} \] for any \(r \geqslant 0\), any measurable set \(A\) in \(\left(\mathbb{R}^{k}\right)^{n}\) such that \(\mu^{n}\left( A \right)\geqslant 1/2\). Here \(B\) is the unit ball in \(\left(\mathbb{R}^{k}\right)^{n}\). The standard Gaussian measure on \(\mathbb{R}\) is such a measure. The author shows that this property can be characterized in terms of transportation-cost inequalities. One of the applications is a new proof of the result of \textit{F. Otto} and \textit{C. Villani} [J. Funct. Anal. 173, No. 2, 361--400 (2000; Zbl 0985.58019)], as well as equivalence of the Poincaré to a dimension free concentration property and a transportation inequality. The proof relies on large deviations methods.