Fuzzy Rule-Based Design of Evolutionary Algorithm for Optimization.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Evolutionary Computation

Background:

• Multi-operator and multi-method evolutionary algorithms often show inconsistent performance across diverse problems.
• Current algorithm design relies heavily on trial-and-error, lacking a systematic approach.
• A need exists for more robust and consistently performing optimization techniques.

Purpose of the Study:

• To propose a novel cooperative framework for evolutionary algorithms.
• To enhance optimization performance through a rule-based cooperative strategy.
• To improve the consistency and effectiveness of evolutionary computation methods.

Main Methods:

• Developed a cooperative strategy where two algorithms form a team, utilizing fuzzy rules based on solution quality and population diversity.
• Implemented two subpopulations, dynamically emphasizing the better-performing one.
• Incorporated a learning mechanism where inferior algorithms learn from superior ones, with fine-tuning in later stages.

Main Results:

• The proposed cooperative evolutionary algorithm demonstrated a high success rate on benchmark datasets (CEC2014, CEC2013, CEC2005, and 12 classical problems).
• The algorithm consistently outperformed traditional single-operator-based algorithms.
• It also showed superior performance compared to various state-of-the-art algorithms in the literature.

Conclusions:

• The proposed fuzzy rule-based cooperative approach offers a significant improvement over existing evolutionary algorithms.
• This method provides a more systematic and effective way to design cooperative optimization strategies.
• The enhanced performance and consistency suggest broader applicability in complex optimization tasks.