A scatter learning particle swarm optimization algorithm for multimodal problems.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Swarm Intelligence

Background:

• Particle Swarm Optimization (PSO) is effective for function optimization.
• PSO performance relies on exemplar characteristics, including fitness and distribution.
• Addressing multimodal problems requires specialized optimization strategies.

Purpose of the Study:

• To introduce a novel PSO variant, the Scatter Learning PSO Algorithm (SLPSOA).
• To enhance PSO's ability to handle multimodal optimization problems.
• To improve the search efficiency and solution quality of PSO.

Main Methods:

• Developed SLPSOA with an exemplar pool (EP) of scattered, high-quality solutions.
• Particles select exemplars from the EP using a roulette wheel rule.
• Integrated Solis and Wets' algorithm for local search to refine solutions.

Main Results:

• SLPSOA demonstrated effectiveness on 16 benchmark functions.
• Compared to six existing PSO algorithms, SLPSOA showed superior performance.
• The proposed algorithm successfully prevented premature convergence.

Conclusions:

• SLPSOA is an efficient variant for multimodal function optimization.
• The exemplar pool and local search mechanisms enhance solution-finding capabilities.
• SLPSOA offers competitive solutions and robust performance in complex optimization tasks.