A double association-based evolutionary algorithm for many-objective optimization.

Area of Science:

• Computational intelligence
• Optimization algorithms
• Evolutionary computation

Background:

• Many-objective optimization problems (MaOPs) present significant computational challenges.
• Existing evolutionary algorithms struggle with maintaining diversity and convergence in high-dimensional objective spaces.

Purpose of the Study:

• To propose a novel double association-based evolutionary algorithm (DAEA) for solving MaOPs.
• To enhance the exploration capabilities and solution quality evaluation in many-objective optimization.

Main Methods:

• Introduction of a double association strategy that links solutions to subspaces, including empty ones, to promote exploration.
• Development of a refined quality evaluation scheme measuring convergence, global diversity, and local diversity.
• Implementation of a dynamic penalty coefficient to balance convergence and diversity.

Main Results:

• DAEA demonstrated high competitiveness against five state-of-the-art algorithms on benchmark MaOPs with up to 20 objectives.
• The proposed double association strategy improved the exploration of unknown areas.
• The refined quality evaluation scheme provided a more nuanced assessment of solution quality.

Conclusions:

• DAEA offers a promising approach for effectively solving many-objective optimization problems.
• The novel strategies enhance the algorithm's ability to find diverse and well-converged solutions.
• DAEA represents a significant advancement in the field of evolutionary computation for MaOPs.