Several Separations Based on a Partial Boolean Function

TL;DR

This paper constructs a partial Boolean function demonstrating large certificate complexity at specific inputs, leading to several significant complexity separations in query and communication complexity.

Contribution

It introduces a partial Boolean function with high certificate complexity at certain inputs, enabling new and improved complexity separations in various computational models.

Findings

Established a function with certificate complexity nearly quadratic in its degree.

Achieved the first known separation between certificate complexity and polynomial degree since 1995.

Provided near-optimal lower bounds for communication complexity and graph theory problems.

Abstract

We show a partial Boolean function $f$ together with an input $x\in f^{-1}\left(*\right)$ such that both $C_{\bar{0}}\left(f,x\right)$ and $C_{\bar{1}}\left(f,x\right)$ are at least $C\left(f\right)^{2-o\left(1\right)}$. Due to recent results by Ben-David, G\"{o}\"{o}s, Jain, and Kothari, this result implies several other separations in query and communication complexity. For example, it gives a function $f$ with $C(f)=\Omega(deg^{2-o\left(1\right)}(f))$ where $C$ and $deg$ denote certificate complexity and polynomial degree of $f$. (This is the first improvement over a separation between $C(f)$ and $deg(f)$ by Kushilevitz and Nisan in 1995.) Other implications of this result are an improved separation between sensitivity and polynomial degree, a near-optimal lower bound on conondeterministic communication complexity for Clique vs. Independent Set problem and a near-optimal lower bound on complexity of Alon--Saks--Seymour problem in graph theory.