Edge-set reduction to efficiently solve the graph partitioning problem with the genetic algorithm

TL;DR

This paper explores how reducing the edge set in graph partitioning problems can improve the efficiency of genetic algorithms, especially for large dense instances, by analyzing different levels of edge reduction.

Contribution

It introduces a method of edge-set reduction to optimize the chromosome size in edge-based genetic algorithms for GPP, enhancing computational efficiency.

Findings

Reduced edge sets improve GA efficiency on large dense graphs

Different levels of edge reduction impact solution quality and computational time

Optimal reduction levels balance solution accuracy and performance

Abstract

The graph partitioning problem (GPP) is among the most challenging models in optimization. Because of its NP-hardness, the researchers directed their interest towards approximate methods such as the genetic algorithms (GA). The edge-based GA has shown promising results when solving GPP. However, for big dense instances, the size of the encoding representation becomes too huge and affects GA's efficiency. In this paper, we investigate the impact of modifying the size of the chromosomes on the edge based GA by reducing the GPP edge set. We study the GA performance with different levels of reductions, and we report the obtained results.