Business Intelligence

R-MAT: A Recursive Model for Graph Mining

Free registration required

Executive Summary

The goal of this paper was to create a simple, parsimonious graph model to describe real graphs. The R-MAT model is exactly a step in this direction: the authors illustrate experimentally that several, diverse real graphs can be well approximated by an R-MAT model with the appropriate choice of parameters. Moreover, they propose a list of natural tests which hold for a variety of real graphs: matching the power-law/DGX distribution for the in- and out-degree; the hop-plot and the diameter of the graph; the singular value distribution; the values of the first singular vector ("Google-score"); and the "Stress" distribution over the edges of the graph.

  • Format: PDF
  • Size: 336.24 KB