Recycling Solutions for Vertex Coloring Heuristics

Abstract

The vertex coloring problem is a well-known NP-hard problem and has many applications in operations research and in scheduling. A conventional approach to the problem solves the k-colorability problem iteratively, decreasing k one by one. Whether a heuristic algorithm finds a legal k-coloring quickly or not is largely affected by an initial solution. We highlight a simple initial solution generator, which we call the recycle method, which makes use of the legal (k+1)-coloring that has been found. An initial solution generated by the method is expected to guide a heuristic algorithm to find a legal k-coloring more quickly than conventional methods, as demonstrated by experimental studies. The results suggest that the recycle method should be used as the standard initial solution generator for both local search algorithms and modern hybrid methods.