Log-concavity and strong log-concavity: a review (Q485901)

This survey article discusses the property of log-concavity in both a continuous and discrete setting, as well as several related notions (such as strong log-concavity, ultra-log-concavity, and their relationship to other such ideas available in the literature). The presentation covers both the univariate and multivariate settings. A large part of the article is devoted to a discussion of the preservation of these log-concavity properties under operations such as convolution, formation of products, marginalization, affine transformations, and distributional limits. This includes a detailed discussion of a monotonicity result due to \textit{B. Efron} [Ann. Math. Stat. 36, 272--279 (1965; Zbl 0134.36704)], and its relationship to the preservation of log-concavity under convolution in the univariate case. A new proof of Efron's theorem, using asymmetric Brascamp-Lieb inequalities, is given. The authors also highlight connections of log-concavity to other areas of mathematics, probability and statistics, including Poincaré, Sobolev and log-Sobolev inequalities, optimal transport, concentration of measure, machine learning, compressed sensing, Laplace approximations, MCMC algorithms, and convex geometry. This article serves as a useful starting point for the vast literature available on log-concavity and related notions, which should prove helpful for researchers at all levels.