Clustering-Based Selection for Evolutionary Many-Objective Optimization

Roman Denysiuk^{1, 3}, Lino Costa^{2, 3}, and Isabel Espírito Santo^{2, 3}

¹Algoritmi R&D Center, University of Minho, Braga, Portugal
roman.denysiuk@algoritmi.uminho.pt

²Department of Production and Systems Engineering, University of Minho, Braga, Portugal
lac@dps.uminho.pt
iapinho@dps.uminho.pt

³University of Minho, Braga, Portugal

Abstract. This paper discusses a selection scheme allowing to employ a clustering technique to guide the search in evolutionary many-objective optimization. The underlying idea to avoid the curse of dimensionality is based on transforming the objective vectors before applying a clustering and the selection of cluster representatives according to the distance to a reference point. The experimental results reveal that the proposed approach is able to effectively guide the search in high-dimensional objective spaces, producing highly competitive performance when compared with state-of-the-art algorithms.

LNCS 8672, p. 538 ff.

Full article in PDF | BibTeX

lncs@springer.com
© Springer International Publishing Switzerland 2014