Log-Sobolev Inequalities and Concentration
In this chapter, we give an introduction to log-Sobolev inequalities and their use in deriving concentration of measure results. This is a third important methodology for concentration of measure (the other two being martingales and transportation cost) and it appears to be the most powerful of the three.
Given a probability space (Ω, P) and a function f : Ω → R, define the entropy of f by
By Jensen’s inequality applied to the convex function ψ(x) := x log x, EntP (f ) ≥ 0 for any f.
A logarithmic Sobolev inequality or just log-Sobolev inequality bounds EntP [f], ...