EvolutionarY Construction of Graph-Coloring Programs using Genetic Programming
Sarit Chantasuban, Ekawit Nantajeewarawat · 2001
Genetic programming (GP) is a novel paradigm that simulates the way of solving problems by nature according to Darwin's theory of fitness-driven natural selection. Instead of using bit strings as in genetic algorithm (GA), GP uses tree structures as its computing structures. As computer programs can be represented as trees, GP has been employed as a method of generating computer programs. In the work reported in this paper, GP is applied to the graph-coloring problem, an NP-complete problem which is an abstraction of many real-world practical problems, with expectation of constructing computer programs that are capable of computing approximations of the optimal solutions to many instances of the problem. The resulting computer programs are analyzed and their performance is compared with two existing commonly used approximation algorithms for graph coloring, i.e., the sequ€ntial coloring algorithms with random coloring order and with maximal-to-minimal-degree coloring order.