Ranking-based hierarchical random mutation in differential evolution.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Evolutionary Computation

Background:

• Differential evolution (DE) is a widely used optimization algorithm.
• Existing DE variants face challenges in balancing global exploration and local exploitation.
• Improving DE performance requires novel mutation strategies and parameter adaptation.

Purpose of the Study:

• To propose a novel algorithm, ranking-based hierarchical random mutation in differential evolution (RHRMDE), to enhance DE performance.
• To introduce a hierarchical mutation mechanism that differentiates strategies for non-inferior and inferior groups.
• To develop an adaptive control parameter strategy considering problem and individual differences.

Main Methods:

• Implemented a hierarchical random mutation mechanism applying "DE/rand/1" and its variant.
• Assigned "DE/rand/1" with global characteristics to the non-inferior group for exploration.
• Utilized an improved "DE/rand/1" with elite characteristics for the inferior group to enhance exploitation.
• Incorporated adaptive control parameter tuning based on problem and individual differences.

Main Results:

• The RHRMDE algorithm demonstrated superior performance compared to five DE variants.
• RHRMDE outperformed five non-DE algorithms on 32 universal benchmark functions.
• The hierarchical mutation strategy effectively balanced global exploration and local exploitation.

Conclusions:

• RHRMDE significantly improves the performance of differential evolution algorithms.
• The proposed hierarchical mutation and adaptive parameter control are effective enhancements.
• RHRMDE offers a robust and efficient approach for complex optimization problems.