Deception in ant colony optimization

Abstract

The search process of a metaheuristic is sometimes misled. This may be caused by features of the tackled problem instance, by features of the algorithm, or by the chosen solution representation. In the field of evolutionary computation, the first case is called deception and the second case is referred to as bias. In this work we formalize the notions of deception and bias for ant colony optimization. We formally define first order deception in ant colony optimization, which corresponds to deception as being described in evolutionary computation. Furthermore, we formally define second order deception in ant colony optimization, which corresponds to the bias introduced by components of the algorithm in evolutionary computation. We show by means of an example that second order deception is a potential problem in ant colony optimization algorithms. © 2004 Springer-Verlag.