Distributed Maximal Independent Set on Scale-Free Networks

TL;DR

This paper introduces a faster distributed algorithm for computing maximal independent sets on scale-free networks, leveraging properties of inhomogeneous random graphs with power-law degree distributions.

Contribution

It presents a novel algorithm with improved time complexity for MIS on scale-free networks and analyzes structural properties like arboricity and degeneracy.

Findings

The induced subgraph with high-degree vertices forms a scale-free network with exponent 2.

The new algorithm achieves an MIS in O(log n / log log n) time, faster than previous O(log n).

On networks with exponent ≥ 3, arboricity and degeneracy are bounded by 2^{log^{1/3} n}.

Abstract

The problem of distributed maximal independent set (MIS) is investigated on inhomogeneous random graphs with power-law weights by which the scale-free networks can be produced. Such a particular problem has been solved on graphs with $n$ vertices by state-of-the-art algorithms with the time complexity of $O(\log{n})$. We prove that for a scale-free network with power-law exponent $\beta > 3$, the induced subgraph is constructed by vertices with degrees larger than $\log{n}\log^{*}{n}$ is a scale-free network with $\beta' = 2$, almost surely (a.s.). Then, we propose a new algorithm that computes an MIS on scale-free networks with the time complexity of $O(\frac{\log{n}}{\log{\log{n}}})$ a.s., which is better than $O(\log{n})$. Furthermore, we prove that on scale-free networks with $\beta \geq 3$, the arboricity and degeneracy are less than $2^{log^{1/3}n}$ with high probability (w.h.p.). Finally, we prove that the time complexity of finding an MIS on scale-free networks with $\beta\geq 3$ is $O(log^{2/3}n)$ w.h.p.