Automatic configuration of multi-objective ACO algorithms

Abstract

In the last few years a significant number of ant colony optimization (ACO) algorithms have been proposed for tackling multi-objective optimization problems. In this paper, we propose a software framework that allows to instantiate the most prominent multi-objective ACO (MOACO) algorithms. More importantly, the flexibility of this MOACO framework allows the application of automatic algorithm configuration techniques. The experimental results presented in this paper show that such an automatic configuration of MOACO algorithms is highly desirable, given that our automatically configured algorithms clearly outperform the best performing MOACO algorithms that have been proposed in the literature. As far as we are aware, this paper is also the first to apply automatic algorithm configuration techniques to multi-objective stochastic local search algorithms. © 2010 Springer-Verlag Berlin Heidelberg.