Objective contribution decomposition method and multi-population optimization strategy for large-scale multi-objective optimization problems

As problem complexity and data dimensionality increase, practical problems tend to have a large number of variables. Large-scale multi -objective optimization problems (LSMOPs) are a complex and difficult class of problems with numerous decision variables to be optimized. In addressing these problems using traditional evolutionary algorithms, achieving convergence is often difficult. The decomposition of decision variables and optimization is a more intuitive way to deal with this problem, which reduces the size of the problem to some extent. Thus, in this study, an objective contribution decomposition method and multi -population optimization strategy are proposed for handling LSMOPs. First, the decision variables are decomposed according to their contribution to the objectives. Further decomposition is performed based on the interactions among the decision variables. Then, multiple populations are generated to optimize the variable group and determine the information interaction among the subpopulations. In experimental results, the performance of the proposed algorithm was competitive in comparisons with six state-of-the-art algorithms.