A study of evaluation functions for the graph K-coloring problem

The evaluation or fitness function is a key component of any heuristic search algorithm. This paper introduces a new evaluation function for the well-known graph K-coloring problem. This function takes into account not only the number of conflicting vertices, but also inherent information related to the structure of the graph. To assess the effectiveness of this new evaluation function, we carry out a number of experiments using a set of DIMACS benchmark graphs. Based on statistic data obtained with a parameter free steepest descent, we show an improvement of the new evaluation function over the classical one.