Preference-driven co-evolutionary algorithms show promise for many-objective optimisation

Abstract

The simultaneous optimisation of four or more conflicting objectives is now recognised as a challenge for evolutionary algorithms seeking to obtain full representations of trade-off surfaces for the purposes of a posteriori decision-making. Whilst there is evidence that some approaches can outperform both random search and standard Pareto-based methods, best-in-class algorithms have yet to be identified. We consider the concept of co-evolving a population of decision-maker preferences as a basis for determining the fitness of competing candidate solutions. The concept is realised using an existing co-evolutionary approach based on goal vectors. We compare this approach and a variant to three realistic alternatives, within a common optimiser framework. The empirical analysis follows current best practice in the field. As the number of objectives is increased, the preference-driven co-evolutionary approaches tend to outperform the alternatives, according to the hypervolume indicator, and so make a strong claim for further attention in many-objective studies. © 2011 Springer-Verlag.