From concentration to logarithmic Sobolev and Poincaré inequalities (Q629718): Difference between revisions

Suppose that \(\left( M, d \right)\) is a smooth Riemannian manifold, \(\mu\) a probability measure on \(M\). If \(\mu\) satisfies the logarithmic Sobolev inequality with a constant \(C\), then it also satisfies the following Gaussian concentration property: for any set \(A\) with \(\mu\left( A \right) \geqslant 1/2\), \[ \mu\left( \left\{ x \in M: d\left( x, A \right)\leqslant r\right\}\right)\geqslant 1-e^{-\frac{r^{2}}{C}} \] for any \(r\geqslant 0\). The converse is also true under certain conditions on the concentration constants and the lower bound on the Ricci curvature. The current article deals with the converse for non-smooth spaces such as length spaces (also known as path metric spaces). The curvature condition on this space is defined in the sense of Lott-Sturm-Villani. A number of other functional inequalities (e.g., Poincaré and Talagrand's transport inequalities) are discussed as well.