Preference-inspired co-evolutionary algorithm using weights for many-objective optimization

Decomposition based approaches are known to perform well on many-objective problems when a suitable set of weights is provided. However, providing a suitable set of weights \textit{a priori} is difficult. This study proposes a novel algorithm: preference-inspired co-evolutionary algorithm using weights (PICEA-w), which co-evolves a set of weights with the usual population of candidate solutions during the search process. The co-evolution enables suitable sets of weights to be constructed along the optimization process, thus guiding the candidate solutions toward the Pareto optimal front. Experimental results show PICEA-w performs better than algorithms embedded with random or uniform weights.