Set-based many-objective optimization guided by preferred regions

Abstract

Set-based evolutionary optimization based on the performance indicators is one of the effective methods to solve many-objective optimization problems; however, previous researches didn’t make full use of the preference information of a high-dimensional objective space to guide the evolution of a population. In this study, we propose a set-based many-objective evolutionary optimization algorithm guided by preferred regions. In the mode of set-based evolution, the proposed method dynamically determines a preferred region of the high-dimensional objective space, designs a selection strategy on sets by combining Pareto dominance relation on sets with the above preferred region, and develops a crossover operator on sets guided by the above preferred region to produce a Pareto front with superior performance. The proposed method is applied to four benchmark many-objective optimization problems, and the experimental results empirically demonstrate its effectiveness.