Parallel and Distributed MOEA/D with Virtual Overlapping Zone and Exclusively Evaluated Mating

We've introduced a technique that establishes a virtual Overlapping zone between partitions. This method facilitates a new way for individuals to both mating and migration by assessing individual populations within this zone. We employ weight vectors from neighboring partitions for this evaluation. This approach is tailored for distributed memory-centric, large-scale parallel and distributed computation, specifically for MOEA/D, which is a decomposition-based evolutionary multi-objective optimization algorithm. In this paper, to further speed up, we verify the effectiveness of the proposed method when without migration. For assessment, we utilized a two-objective constrained knapsack problem. When comparing our proposed method to the standard single-core execution and the No-Overlapping standard parallel MOEA/D, we found that our approach enhances the diversity in solution searches. Additionally, it reduces execution time and bolsters the precision of the search results. Compared to the Overlap method that migrates, the decrease in Hypervolume value is slight, but it shows that it is effective in shortening the execution time.