A remark on Bourgain’s distributional inequality on the Fourier spectrum of Boolean functi.pdf

A remark on Bourgain’s distributional inequality on the Fourier spectrum of Boolean functions Hamed Hatami Department of Computer Science University of Toronto E-mail: hamed@cs.toronto.edu Submitted: September 24, 2005; Accepted: May 2, 2006; Published: May 5, 2006 Abstract Bourgain’s theorem says that under certain conditions a function f : {0, 1}n → {0, 1} can be approximated by a 2 function g which depends only on a small number of variables. By following his proof we obtain a generalization for the case that there is a nonuniform product measure on the domain of f . 1 Introduction Fix 0 α 1. Consider the measure space (Fn , µ ) where µ is the product measure 2 α α defined as µ (x) = α|x | (1 − α)n−|x | with |x | = n x . Let f : (Fn , µ ) → {0, 1} be a α i=1 i 2 α Boolean function. We want to show that under certain conditions f can be approximated by a function g which depends only on a small number of variables. More formally, there exist indices 1 ≤ i1 . . . im ≤ n such g (x) = g (y), if xij = yij for every 1 ≤ j ≤ m. Moreover by g approximates f we mean that 2 f − g ≤ . (1) 2 Here m and are parameters which depend on each other and the con