Some techniques to deal with many-objective problems

Abstract

Although Multi-objective Evolutionary Algorithms (MOEAs) have successfully been used in a wide range of real-world problems, recent studies have shown that they have scalability limitations in problems with a large number of objectives. This paper present two contributions to deal with those limitations. In one contribution we developed an objective reduction technique which can be used during the search or in the decision making process. The other contribution presents a comparison study of some preference relations designed for problems with a large number of objectives. The experimental results show that both approaches successfully deal, to a great extent, with such scalability limitations.