# A parallel implementation of the Synchronised Louvain method

## TL;DR

This paper presents a parallel implementation of the Synchronised Louvain method for community detection, significantly improving computation time on large graphs by leveraging OpenMP for parallelization.

## Contribution

The paper introduces a parallel version of Browet's Synchronised Louvain algorithm, demonstrating substantial speedup and performance improvements on large-scale networks.

## Key findings

- Achieved significant speedup over sequential implementation
- Parallel algorithm performs well on large graphs
- Results align with theoretical predictions of Amdahl's law

## Abstract

Community detection in networks is a very actual and important field of research with applications in many areas. But, given that the amount of processed data increases more and more, existing algorithms need to be adapted for very large graphs. The objective of this project was to parallelise the Synchronised Louvain Method, a community detection algorithm developed by Arnaud Browet, in order to improve its performances in terms of computation time and thus be able to faster detect communities in very large graphs. To reach this goal, we used the API OpenMP to parallelise the algorithm and then carried out performance tests. We studied the computation time and speedup of the parallelised algorithm and were able to bring out some qualitative trends. We obtained a great speedup, compared with the theoretical prediction of Amdahl law. To conclude, using the parallel implementation of the algorithm of Browet on large graphs seems to give good results, both in terms of computation time and speedup. Further tests should be carried out in order to obtain more quantitative results.