Hybridizing Niching, Particle Swarm Optimization, and Evolution Strategy for Multimodal Optimization.

Area of Science:

• Computational intelligence
• Optimization algorithms
• Machine learning

Background:

• Multimodal optimization problems (MMOPs) often feature multiple optimal solutions, posing challenges for standard algorithms.
• Existing methods may struggle to efficiently identify and maintain diverse optimal solutions, risking convergence to a single peak.

Purpose of the Study:

• To introduce a novel population division technique, nearest-better-neighbor clustering (NBNC), for MMOPs.
• To develop and evaluate a new algorithm, NBNC-PSO-ES, integrating NBNC with particle swarm optimization (PSO) and covariance matrix adaption evolution strategy (CMA-ES).

Main Methods:

• Proposed nearest-better-neighbor clustering (NBNC) for population division to prevent multiple species converging to the same optimum.
• Constructed raw species by linking individuals to their better neighbors, then merged dominated raw species for final population structuring.
• Developed NBNC-PSO-ES, combining PSO's exploration with CMA-ES's exploitation capabilities.

Main Results:

• NBNC-PSO-ES was tested against state-of-the-art algorithms on benchmark MMOPs.
• Experimental results indicated that NBNC-PSO-ES outperformed comparative algorithms in performance.

Conclusions:

• The proposed NBNC method effectively reduces the risk of multiple species converging to the same peak in MMOPs.
• NBNC-PSO-ES offers a promising approach for solving multimodal optimization problems, demonstrating enhanced performance.