Regularization under diffusion and anti-concentration of the information content
arXiv:1410.3887 · doi:10.1215/00127094-2017-0048
Abstract
Under the Ornstein-Uhlenbeck semigroup , any non-negative measurable exhibits a uniform tail bound better than that implied by Markov's inequality and conservation of mass: For every , and , \[ γ_n\left(\left\{x \in \mathbb R^n : U_t f(x) > α\int f\,dγ_n\right\}\right) \leq C(t) \frac{1}α \sqrt{\frac{\log \log α}{\log α}}\] where is the -dimensional Gaussian measure and is a constant depending only on . 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 is {\em semi-log-convex} in the sense that for some , for all , the eigenvalues of are at least . Then satisfies a tail bound asymptotically better than that implied by Markov's inequality.
arXiv commentThe bound is improved and the proof have been significantly simplified
References in corpus (3)
Cited by in corpus (10)
- Neural Stochastic Differential Equations: Deep Latent Gaussian Models in the Diffusion Limit
- Theoretical guarantees for sampling and inference in generative models with latent diffusions
- Stability of the Shannon-Stam inequality via the Föllmer process
- Deviation inequalities for convex functions motivated by the Talagrand conjecture
- Log-Hessian and Deviation Bounds for Markov Semi-Groups, and Regularization Effect in
- Regularization in for the Ornstein--Uhlenbeck semigroup
- Regularized Brascamp--Lieb inequalities
- Stability of Talagrand's Gaussian transport-entropy inequality via the Föllmer process
- -optimal transport of anisotropic log-concave measures and exponential convergence in Fisher's infinitesimal model
- Semi-log-convexity of queues on