Differential Evolution for Multimodal Optimization With Species by Nearest-Better Clustering.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Machine Learning

Background:

• Multimodal optimization problems (MMOPs) require identifying multiple optimal solutions for real-world decision-making.
• Balancing global exploration and local exploitation is crucial for effectively solving MMOPs.

Purpose of the Study:

• To propose a novel algorithm, FBK-DE, for efficiently solving multimodal optimization problems.
• To enhance the ability to explore the global search space and exploit multiple optimal regions.

Main Methods:

• Developed FBK-DE based on differential evolution (DE).
• Utilized nearest-better clustering (NBC) to partition populations into species with minimum size constraints.
• Implemented a species balance strategy to regulate population distribution across species.
• Introduced two keypoint-based mutation operators (DE/keypoint/1 and DE/keypoint/2) for enhanced species evolution.

Main Results:

• FBK-DE was evaluated on 20 benchmark functions.
• Performance was compared against 15 state-of-the-art multimodal optimization algorithms.
• FBK-DE demonstrated competitive results, showing its efficacy in handling MMOPs.

Conclusions:

• The proposed FBK-DE algorithm effectively addresses the challenges of multimodal optimization.
• The formulation, balance, and keypoint strategies contribute to improved performance in identifying multiple optima.
• FBK-DE offers a competitive alternative for solving complex real-world optimization tasks.