The Communication Complexity of Local Search

TL;DR

This paper investigates the communication complexity of finding local maxima in graphs and demonstrates that such problems require polynomial to exponential communication, with implications for game theory and auctions.

Contribution

It establishes polynomial lower bounds for local search communication complexity on various graphs and provides tight bounds for hypercubes, also applying these results to potential games and auctions.

Findings

Finding local maxima in certain graphs requires polynomial communication.

Optimal bound of  bits for hypercubes.

Exponential communication complexity for Nash equilibria and combinatorial auctions.

Abstract

We study the following communication variant of local search. There is some fixed, commonly known graph $G$. Alice holds $f_A$ and Bob holds $f_B$, both are functions that specify a value for each vertex. The goal is to find a local maximum of $f_A+f_B$ with respect to $G$, i.e., a vertex $v$ for which $(f_A+f_B)(v)\geq (f_A+f_B)(u)$ for every neighbor $u$ of $v$. Our main result is that finding a local maximum requires polynomial (in the number of vertices) bits of communication. The result holds for the following families of graphs: three dimensional grids, hypercubes, odd graphs, and degree 4 graphs. Moreover, we provide an \emph{optimal} communication bound of $\Omega(\sqrt{N})$ for the hypercube, and for a constant dimensional greed, where $N$ is the number of vertices in the graph. We provide applications of our main result in two domains, exact potential games and combinatorial auctions. First, we show that finding a pure Nash equilibrium in $2$-player $N$-action exact potential games requires polynomial (in $N$) communication. We also show that finding a pure Nash equilibrium in $n$-player $2$-action exact potential games requires exponential (in $n$) communication. The second domain that we consider is combinatorial auctions, in which we prove that finding a local maximum in combinatorial auctions requires exponential (in the number of items) communication even when the valuations are submodular. Each one of the results demonstrates an exponential separation between the non-deterministic communication complexity and the randomized communication complexity of a total search problem.