Hypercontractive Inequality for Pseudo-Boolean Functions of Bounded Fourier Width
Abstract
A function f:\ \-1,1\n→ R is called pseudo-Boolean. It is well-known that each pseudo-Boolean function f can be written as f(x)=ΣI∈ Ff(I)I(x), where F⊂eq \I:\ I⊂eq [n]\, [n]=\1,2,...,n\, and I(x)=Πi∈ Ixi and f(I) are non-zero reals. The degree of f is \|I|:\ I∈ F\ and the width of f is the minimum integer such that every i∈ [n] appears in at most sets in F. For i∈ [n], let xi be a random variable taking values 1 or -1 uniformly and independently from all other variables xj, j≠ i. Let x=(x1,...,xn). The p-norm of f is ||f||p=( E[|f(x)|p])1/p for any p 1. It is well-known that ||f||q ||f||p whenever q> p 1. However, the higher norm can be bounded by the lower norm times a coefficient not directly depending on f: if f is of degree d and q> p>1 then ||f||q (q-1p-1)d/2||f||p. This inequality is called the Hypercontractive Inequality. We show that one can replace d by in the Hypercontractive Inequality for each q> p 2 as follows: ||f||q ((2r)!r-1)1/(2r)||f||p, where r= q/2. For the case q=4 and p=2, which is important in many applications, we prove a stronger inequality: ||f||4 (2+1)1/4||f||2.
Turn this paper into a lesson
ArcXiv compiles a structured reading guide from this paper's metadata: plain-English importance, contributions, prerequisite concepts, which sections to read first, flashcards, and a quiz. Grounded in the abstract, never invented.