Transport proofs of weighted Poincaré inequalities for log-concave distributions (Q502860): Difference between revisions

The paper highlights new versions of weighted Poincaré inequalities. Let us associate to an arbitrary \(n\)-dimensional log-concave random vector \(X=\left(X_1,\cdots,X_n\right)\) the ``centred'' \(n\)-dimensional random vector \(\overline{X}\) with the components \(\overline{X}_i=X_i-\mathbb{E}\left[X_i|X_1,\cdots,X_{i-1}\right],\;i=1,\cdots,n\). The main result of the paper states that there exists an universal constant \(a>0\) such that the Poincaré type inequality \(\text{Var}\left(f\left(\overline{X}\right)\right)\leq a\sum\limits_{i=1}^n\mathbb{E}\left[\mathbb{E}\left[\overline{X}^2_i|\overline{X}_i|\overline{X}_i,\cdots,\overline{X}_{i-1}\right]\partial_if\left(\overline{X}\right)^2\right]\) holds for all locally-Lipschitz functions \(f:\mathbb{R}^n\to \mathbb{R}\) which belong to \(\mathbb{L}_2\left(\overline{X}\right)\). The proof is based on a novel transport-entropy inequality involving the optimal transport cost and the Kullback-Leibler distance. Many interesting consequences and refined observations are also proposed. This work extends recent similar studies due to \textit{B. Klartag} [Invent. Math. 168, No. 1, 91--131 (2007; Zbl 1144.60021)] and to \textit{F. Barthe} and the first author [Proc. Lond. Math. Soc. (3) 106, No. 1, 33--64 (2013; Zbl 1281.60020)].