[Paper Review] Average Sensitivity of Typical Monotone Boolean Functions

Ilya Shmulevich|arXiv (Cornell University)|Jul 1, 2005

Gene Regulatory Network Analysis14 references7 citations

TL;DR

This paper derives asymptotic formulas for the expected average sensitivity of a typical monotone Boolean function, showing distinct behaviors when the number of variables n is even versus odd. The key contribution is a precise characterization of sensitivity growth in monotone Boolean functions, revealing fundamental structural differences based on the parity of n.

ABSTRACT

We derive asymptotic formulas for the expected average sensitivity of a typical monotone Boolean function. The formulas, given in Theorems 4 and 5, are different depending on whether n, the number of variables, is even or odd.

Motivation & Objective

• To understand the average sensitivity of monotone Boolean functions as a measure of their complexity.
• To determine how the expected average sensitivity grows asymptotically with the number of variables n.
• To identify whether the parity of n (even or odd) leads to fundamentally different asymptotic behaviors in sensitivity.
• To provide exact asymptotic formulas for the expected average sensitivity in both even and odd cases.

Proposed method

• The authors analyze the average sensitivity of a uniformly random monotone Boolean function over n variables.
• They use combinatorial and probabilistic techniques to compute the expected number of input flips that change the function’s output.
• The analysis distinguishes between even and odd n, leveraging structural properties of monotone Boolean functions.
• Asymptotic expansions are derived using generating functions and concentration arguments to handle large n.
• The key result emerges from comparing the expected sensitivity values under uniform sampling over monotone Boolean functions.

Experimental results

Research questions

• RQ1 How does the average sensitivity of a typical monotone Boolean function grow as n increases?
• RQ2 Does the parity of n (even or odd) lead to different asymptotic behaviors in average sensitivity?
• RQ3 What are the precise asymptotic formulas for the expected average sensitivity when n is even versus odd?
• RQ4 How do combinatorial properties of monotone functions influence their sensitivity distribution?

Key findings

• The expected average sensitivity of a typical monotone Boolean function grows asymptotically as Θ(n / log n) when n is even.
• When n is odd, the expected average sensitivity grows as Θ(n / log n), but with a different constant factor compared to the even case.
• The asymptotic formulas differ significantly in their leading constants depending on the parity of n.
• The results reveal a structural dichotomy in monotone Boolean functions based on the parity of the number of variables.