Bio-inspired Information and Communication Technologies. 11th EAI International Conference, BICT 2019, Pittsburgh, PA, USA, March 13–14, 2019, Proceedings

Research Article

A Distribution Control of Weight Vector Set for Multi-objective Evolutionary Algorithms

Download
327 downloads
  • @INPROCEEDINGS{10.1007/978-3-030-24202-2_6,
        author={Tomoaki Takagi and Keiki Takadama and Hiroyuki Sato},
        title={A Distribution Control of Weight Vector Set for Multi-objective Evolutionary Algorithms},
        proceedings={Bio-inspired Information and Communication Technologies. 11th EAI International Conference, BICT 2019, Pittsburgh, PA, USA, March 13--14, 2019, Proceedings},
        proceedings_a={BICT},
        year={2019},
        month={7},
        keywords={Multi-objective optimization Evolutionary computation},
        doi={10.1007/978-3-030-24202-2_6}
    }
    
  • Tomoaki Takagi
    Keiki Takadama
    Hiroyuki Sato
    Year: 2019
    A Distribution Control of Weight Vector Set for Multi-objective Evolutionary Algorithms
    BICT
    Springer
    DOI: 10.1007/978-3-030-24202-2_6
Tomoaki Takagi1,*, Keiki Takadama1,*, Hiroyuki Sato1,*
  • 1: The University of Electro-Communications
*Contact email: tomtkg@uec.ac.jp, keiki@inf.uec.ac.jp, sato@hc.uec.ac.jp

Abstract

For solving multi-objective optimization problems with evolutionary algorithms, the decomposing the Pareto front by using a set of weight vectors is a promising approach. Although an appropriate distribution of weight vectors depends on the Pareto front shape, the uniformly distributed weight vector set is generally employed since the shape is unknown before the search. This work proposes a simple way to control the weight vector distribution appropriate for several Pareto front shapes. The proposed approach changes the distribution of the weight vector set based on the intermediate objective vector in the objective space. A user-defined parameter determines the intermediate objective vector in the static method, and the objective values of the obtained solutions dynamically determine the intermediate objective vector in the dynamic method. In this work, we focus on MOEA/D as a representative decomposition-based multi-objective evolutionary algorithm and apply the proposed static and dynamic methods for it. The experimental results on WFG test problems with different Pareto front shapes show that the proposed static and dynamic methods improve the uniformity of the obtained solutions for several Pareto front shapes and the dynamic method can find an appropriate intermediate objective vector for each Pareto front shape.