Work

Improving selection and reproduction efficiency by exploiting local mini-archives in the decomposition-based evolutionary multi-objective optimization

Year: 2026
Type: article
Abstract: This paper presents novel approaches to improve the efficiency of selection and reproduction in decomposition-based evolutionary multi-objective optimization. The central innovation is the introduction of local mini-archives associated with subproblems. These structures locally preserve the best-contributing solutions and are designed for computational efficiency. Their utility is demonstrated in several ways. First, they help mitigate local convergence deterioration during environmental selection. Second, they enable more efficient parent selection by focusing on local incumbents in the later stages of evolution. Third, the method leverages these mini-archives to perform effective gradient-based adjustments to offspring decision vectors generated by standard crossover and mutation. These developments were rigorously evaluated in both a posteriori and interactive optimization contexts. Interactive optimization, in particular, is more demanding and requires multiple population adaptations within a single optimizer run. In both scenarios, the proposed methods facilitated the construction of higher-quality populations with fewer function evaluations.
Language: English
Open Access status: hybrid
Estimated APC paid: $3,690
OpenAlex ID: W7160390849
FWCI:
Cites: 56
Cited by:
Related to: 10