# Linear Work Generation of R-MAT Graphs

## TL;DR

This paper presents an improved R-MAT graph generation algorithm that reduces work per edge to constant, enabling efficient, parallel graph creation with complex network properties.

## Contribution

It introduces a novel R-MAT generation method that significantly decreases computational work per edge and is highly parallelizable.

## Key findings

- Work per edge reduced to constant
- Algorithm is embarrassingly parallel
- Maintains complex network properties

## Abstract

R-MAT is a simple, widely used recursive model for generating `complex network' graphs with a power law degree distribution and community structure. We make R-MAT even more useful by reducing the required work per edge from logarithmic to constant. The algorithm works in an embarrassingly parallel way.