Computational benchmark study demonstrates improved search performance in large-scale multi-objective optimization problems, suggesting dual-space feedback enhances evolutionary algorithms.
Key Points
To address insufficient selection pressure in large-scale multi-objective optimization by coordinating decision-space and objective-space selection strategies.
Developed MOEA-RLD, an algorithm integrating an adaptive dual-space discriminator, a state-aware recurrent operator selector, a bounded population pool, and angle-based environmental selection.
Evaluated performance using the LSMOP benchmark test suite against five representative large-scale multi-objective evolutionary algorithms and three real-world application problems.
MOEA-RLD demonstrated superior average rankings compared to five representative large-scale evolutionary algorithms across benchmark test settings.
Practical tests across three real-world applications confirmed strong optimization performance alongside specific operational limitations.