The Generalized Differential Evolution 3 (GDE3) is one of the multi-objective evolutionary algorithms. The method has a problem on keeping the diversity of the solutions. In the GDE3, the Crowding Distance (CD) is used for deleting redundantindividuals. If the number of the objectives becomes larger than three, the CD is not suitable for measure the neighborhood relationships among the individuals. This paper proposes the algorithm that uses the Gabriel graph in order to get uniformly distributed solutions for multi-objective problems. Neighborhood relationship among the individuals are represented by the Gabriel graph, and the graph structure is utilized for deleting redundant individuals. As the results of experiments using three-objective problems, our proposed method showed better performance than the conventional GDE3.

2011 IEEE SMC Hiroshima Chapter若手研究会, 2011年7月9日, 広島市立大学