Differential evolution with two-level parameter adaptation.

Area of Science:

• Optimization Algorithms
• Computational Intelligence
• Evolutionary Computation

Background:

• Differential Evolution (DE) algorithm performance is sensitive to mutation strategies and control parameters.
• Existing DE variants often struggle to balance fast convergence with maintaining population diversity.

Purpose of the Study:

• To propose a novel Adaptive Differential Evolution (ADE) algorithm.
• To introduce a new mutation strategy, DE/lbest/1, and a two-level adaptive parameter control scheme.
• To improve the balance between convergence speed and population diversity in DE algorithms.

Main Methods:

• Implemented a DE/lbest/1 mutation strategy using multiple local best individuals.
• Developed a two-level adaptive parameter control: population-level (Fp, CRp) based on optimization states (exploration/exploitation), and individual-level (Fi, CRi) based on fitness and distance to the global best.
• Evaluated ADE on benchmark functions and real-world optimization problems.

Main Results:

• The proposed ADE algorithm demonstrated superior performance compared to four state-of-the-art DE variants across various optimization problems.
• Analysis confirmed the effectiveness of ADE's components, parameter properties, search behavior, and parameter sensitivity.
• ADE showed capability in solving complex real-world optimization tasks.

Conclusions:

• The novel DE/lbest/1 mutation strategy and two-level adaptive parameter control significantly enhance DE performance.
• ADE offers a robust and adaptable solution for optimization problems, outperforming existing methods.
• The adaptive nature of ADE allows it to effectively handle diverse optimization landscapes and real-world challenges.