Low-rank Similarity Measure for Role Model Extraction

TL;DR

This paper introduces a low-rank iterative similarity measure for large networks that effectively identifies node roles based on flow patterns, matching the performance of traditional measures.

Contribution

It proposes a novel low-rank iterative scheme to efficiently compute node similarity for role extraction in large networks.

Findings

Successfully extracts roles in random graphs

Performs comparably to existing pairwise similarity measures

Efficiently scales to very large networks

Abstract

Computing meaningful clusters of nodes is crucial to analyze large networks. In this paper, we present a pairwise node similarity measure that allows to extract roles, i.e. group of nodes sharing similar flow patterns within a network. We propose a low rank iterative scheme to approximate the similarity measure for very large networks. Finally, we show that our low rank similarity score successfully extracts the different roles in random graphs and that its performances are similar to the pairwise similarity measure.