| Abstract: |
Multiobjective optimization in general aims at learning about the problem at hand. Usually the focus lies on objective space properties such as the front shape and the distribution of optimal
solutions. However, structural characteristics in the decision space can also provide valuable insights. In cer- tain applications, it may even be more important to find a structurally diverse
set of close-to-optimal solutions than to identify a set of optimal but structurally similar solutions. Accordingly, multiobjective optimizers are required that are capable of considering both
the objective space quality of a Pareto-set approximation and its diversity in the decision space. Although NSGA, one of the first multiobjective evolution- ary algorithms, explicitly considered
decision space diversity, only a few other studies address that issue. It therefore is an open research question how modern multiobjective evolu- tionary algorithms can be adapted to search for
structurally diverse high-quality Pareto-set approximations. To this end we propose an approach to integrate decision space diversity into hypervolume-based multiobjective search. We present a
modified hypervolume indicator and integrate it into an evolutionary algorithm. The proof-of-principle results show the potential of the approach and indicate further research directions for
structure-oriented multiobjective search. |
