Regularization under diffusion and anticoncentration of the information content

From MaRDI portal




Abstract: Under the Ornstein-Uhlenbeck semigroup Ut, any non-negative measurable f:mathbbRnomathbbR+ exhibits a uniform tail bound better than that implied by Markov's inequality and conservation of mass: For every alphageqe3, and t>0, [ gamma_nleft(left{x in mathbb R^n : U_t f(x) > alpha int f,dgamma_n ight} ight) leq C(t) frac{1}{alpha} sqrt{frac{log log alpha}{log alpha}}] where gamman is the n-dimensional Gaussian measure and C(t) is a constant depending only on t. This confirms positively the Gaussian limiting case of Talagrand's convolution conjecture (1989). This is shown to follow from a more general phenomenon. Suppose that f:mathbbRnomathbbR+ is {em semi-log-convex} in the sense that for some , for all xinmathbbRn, the eigenvalues of abla2logf(x) are at least . Then f satisfies a tail bound asymptotically better than that implied by Markov's inequality.



Cites work







This page was built for publication: Regularization under diffusion and anticoncentration of the information content

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1750377)