Enhanced differential evolution with adaptive strategies for numerical optimization.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Evolutionary Computation

Background:

• Differential evolution (DE) is a powerful global numerical optimization technique.
• Selecting the optimal mutation strategy for DE remains a challenge for specific problems.
• Existing DE variants often require manual parameter tuning or strategy selection.

Purpose of the Study:

• To develop an adaptive strategy adaptation mechanism (SaM) for differential evolution (DE).
• To enhance the performance and robustness of DE algorithms by automating strategy selection.
• To investigate the efficacy of SaM in conjunction with parameter adaptation methods.

Main Methods:

• Proposed a novel strategy adaptation mechanism (SaM) for DE.
• Integrated SaM with the JADE (Just Another DE) variant.
• Evaluated the approach on 20 benchmark scalable optimization problems.
• Tested SaM's ability to adaptively choose suitable strategies and parameters.

Main Results:

• SaM successfully adapted to select more suitable mutation strategies for specific problems.
• The combined SaM-JADE approach demonstrated superior or comparable performance against state-of-the-art DE variants.
• Improvements were observed in both the quality of final solutions and convergence rates.
• The approach proved effective in solving real-world optimization problems.

Conclusions:

• The proposed strategy adaptation mechanism (SaM) significantly enhances differential evolution performance.
• SaM offers a flexible framework for integrating various parameter adaptation techniques.
• This adaptive approach provides a more robust and efficient solution for global numerical optimization.