Applications of Evolutionary Computation to Real-World Multi-Objective Optimization Problems and their Evaluation Criteria

Abstract

In multi-objective optimization problems, there are differences in search difficulty between objective functions. Due to the characteristics of GA, the optimization to difficult objective function(s) does not proceed, and biased search can be performed. However, for the designing, balanced optimization for all objective functions is required. Therefore, in this paper, we propose a method to allocate search resources with priority to difficult objective functions. The purpose of the proposed method is optimizing all objective functions equally regardless of search difficulty of objective functions. We also propose the evaluation criterion for balanced search.