Steady state particle swarm.

Area of Science:

• Computational intelligence
• Optimization algorithms
• Evolutionary computation

Background:

• Particle Swarm Optimization (PSO) is a widely used metaheuristic algorithm.
• Standard PSO can suffer from slow convergence and suboptimal solutions.
• Existing dynamic parameter and neighborhood strategies offer improvements but can be complex.

Purpose of the Study:

• To propose and evaluate a novel steady-state and dynamic update strategy for PSO.
• To enhance the performance and scalability of PSO algorithms.
• To leverage principles from the Bak-Sneppen model of co-evolution for optimization.

Main Methods:

• A new update strategy inspired by the Bak-Sneppen model was developed for PSO.
• The strategy involves updating only the least fit particle and its neighbors.
• Performance was evaluated on unimodal, multimodal, noisy, and rotated benchmark functions.

Main Results:

• The proposed steady-state PSO significantly improved result quality compared to standard and dynamic PSO variants.
• Faster convergence speeds were observed with the new update strategy.
• Sensitivity analysis confirmed performance enhancements across various parameter settings.

Conclusions:

• The Bak-Sneppen-inspired update strategy offers a robust and efficient method for improving PSO performance.
• The algorithm demonstrates consistent behavior and scalability across different dimensions.
• This novel approach provides a valuable advancement in optimization techniques.