Guiding Points Sampling-Based Large-Scale Evolutionary Multi-Objective Optimization
Haoyuan Zeng · 2021 7th International Conference on Computer and Communications (ICCC) · 2021
It is especially hard to search along the correct directions to generate promising offspring solutions in the large-scale decision space by using evolutionary algorithms. To tackle this problem, a guiding points-based sampling strategy for large-scale multi-objective optimization is proposed. First, guiding points are updated by the combination of the historical information and a set of candidate solutions which are the closest to the ideal point in the decision space. Then, a two-stage sampling strategy i.e. convergence-related sampling strategy and diversity-related sampling strategy is performed by the assistance of guiding points. To be specific, the convergence-related sampling strategy leverages the distance relationships between guiding points and candidate solutions to enhance the convergence. Furthermore, the diversity-related sampling strategy considers the distribution relationships between guiding points to promote the diversity for the next generation population. Experimental results on nine large-scale multi-objective optimization benchmark problems show the effectiveness of the proposed algorithm.