Complex Contagions on Configuration Model Graphs with a Power-Law Degree Distribution

Conference paper

DOI: 10.1007/978-3-662-54110-4_32

Part of the Lecture Notes in Computer Science book series (LNCS, volume 10123)
Cite this paper as:
Schoenebeck G., Yu FY. (2016) Complex Contagions on Configuration Model Graphs with a Power-Law Degree Distribution. In: Cai Y., Vetta A. (eds) Web and Internet Economics. WINE 2016. Lecture Notes in Computer Science, vol 10123. Springer, Berlin, Heidelberg

Abstract

In this paper we analyze k-complex contagions (sometimes called bootstrap percolation) on configuration model graphs with a power-law distribution. Our main result is that if the power-law exponent \(\alpha \in (2, 3)\), then with high probability, the single seed of the highest degree node will infect a constant fraction of the graph within time \(O\left( \log ^{\frac{\alpha -2}{3-\alpha }}(n)\right) \). This complements the prior work which shows that for \(\alpha > 3\) boot strap percolation does not spread to a constant fraction of the graph unless a constant fraction of nodes are initially infected. This also establishes a threshold at \(\alpha = 3\).

The case where \(\alpha \in (2, 3)\) is especially interesting because it captures the exponent parameters often observed in social networks (with approximate power-law degree distribution). Thus, such networks will spread complex contagions even lacking any other structures.

We additionally show that our theorem implies that \(\omega (\left( n^{\frac{\alpha -2}{\alpha -1}}\right) \) random seeds will infect a constant fraction of the graph within time \(O\left( \log ^{\frac{\alpha -2}{3-\alpha }}(n)\right) \) with high probability. This complements prior work which shows that \(o\left( n^{\frac{\alpha -2}{\alpha -1}}\right) \) random seeds will have no effect with high probability, and this also establishes a threshold at \(n^{\frac{\alpha -2}{\alpha -1}}\).

Copyright information

© Springer-Verlag GmbH Germany 2016

Authors and Affiliations

  1. 1.University of MichiganAnn ArborUSA

Personalised recommendations