On the model-building issue of multi-objective estimation of distribution algorithms

Abstract

It has been claimed that perhaps a paradigm shift is necessary in order to be able to deal with this scalability issue of multi-objective optimization evolutionary algorithms. Estimation of distribution algorithms are viable candidates for such task because of their adaptation and learning abilities and simplified algorithmics. Nevertheless, the extension of EDAs to the multi-objective domain have not provided a significant improvement over MOEAs. In this paper we analyze the possible causes of this underachievement and propose a set of measures that should be taken in order to overcome the current situation. © 2009 Springer Berlin Heidelberg.