A Diversity-Enhanced Resource Allocation Strategy for Decomposition-Based Multiobjective Evolutionary Algorithm.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Evolutionary Computation

Background:

• Multiobjective evolutionary algorithms (MOEAs) based on decomposition break down complex problems into smaller subproblems.
• Resource allocation (RA) strategies are crucial for efficiently assigning computational resources to these subproblems.
• Current RA schemes primarily rely on aggregated function improvement, potentially neglecting subproblem diversity.

Purpose of the Study:

• To propose a novel diversity-enhanced resource allocation (RA) strategy for decomposition-based MOEAs.
• To improve the performance of MOEAs by dynamically assigning computational resources based on both performance improvement and solution density.

Main Methods:

• Developed a new RA strategy that integrates relative improvement of aggregated function values with solution density around each subproblem.
• Implemented the proposed strategy within a decomposition-based MOEA framework.
• Evaluated the strategy's effectiveness on a suite of challenging benchmark problems.

Main Results:

• The proposed diversity-enhanced RA strategy demonstrated superior performance compared to two existing popular RA strategies.
• Experimental results indicate improved efficiency in tackling complicated multiobjective optimization problems.
• The strategy effectively balances resource allocation by considering both solution convergence and distribution.

Conclusions:

• The proposed diversity-enhanced RA strategy offers a significant advancement for decomposition-based MOEAs.
• Considering solution density alongside performance improvement leads to more effective resource allocation.
• This approach enhances the ability of MOEAs to solve complex multiobjective optimization problems.