A competitive swarm optimizer for large scale optimization.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Swarm Intelligence

Background:

• Particle Swarm Optimization (PSO) is a widely used metaheuristic.
• Large scale optimization problems present significant computational challenges.
• Existing metaheuristics may struggle with balancing exploration and exploitation in high dimensions.

Purpose of the Study:

• To propose a novel Competitive Swarm Optimizer (CSO) for large scale optimization.
• To introduce a new particle update mechanism based on pairwise competition.
• To analyze the convergence, exploration, and exploitation characteristics of CSO.

Main Methods:

• Developed a Competitive Swarm Optimizer (CSO) algorithm.
• Implemented a pairwise competition mechanism for particle updates, where losers learn from winners.
• Provided theoretical convergence proof and empirical analysis of search behavior.
• Benchmarked CSO against five state-of-the-art metaheuristic algorithms on large scale problems.

Main Results:

• CSO demonstrates a good balance between exploration and exploitation.
• Theoretical proof of convergence was established.
• Empirical results show CSO outperforms five state-of-the-art algorithms.
• CSO effectively solves problems with dimensionality up to 5000.

Conclusions:

• The proposed CSO is a simple yet effective algorithm for large scale optimization.
• CSO offers a competitive alternative to existing metaheuristic approaches.
• The pairwise competition mechanism is a promising strategy for swarm intelligence optimization.