Learning Competitive Swarm Optimization

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Swarm Intelligence

Background:

• Particle Swarm Optimization (PSO) is widely applied but struggles with complex problems due to diversity loss and premature convergence.
• Existing PSO variants often fail to adequately address the exploration-exploitation balance in high-dimensional search spaces.

Purpose of the Study:

• To propose a novel Learning Competitive Swarm Optimization (LCSO) algorithm to overcome the limitations of traditional PSO.
• To enhance population diversity and exploration capabilities in complex optimization tasks.

Main Methods:

• The LCSO algorithm divides the swarm into parallel sub-swarms, employing a tournament-based learning mechanism within each sub-swarm.
• A second phase involves inter-sub-swarm information exchange to further enrich the search process.
• The algorithm's performance was evaluated using a suite of test functions.

Main Results:

• The LCSO algorithm demonstrated enhanced particle swarm entropy and an improved search process compared to baseline methods.
• Experimental results showed LCSO significantly outperforms competitive swarm optimizer (CSO), comprehensive particle swarm optimizer (CLPSO), PSO, and other advanced algorithms.
• Statistical analysis confirmed the superior efficiency and effectiveness of the proposed LCSO.

Conclusions:

• The Learning Competitive Swarm Optimization (LCSO) algorithm effectively addresses PSO's limitations in complex multidimensional optimization.
• LCSO offers a robust and statistically superior alternative for enhancing swarm intelligence and solving challenging optimization problems.