Competitive Swarm Optimization for High-Dimensional Search

Summary

The article explains Competitive Swarm Optimizer (CSO), a population-based optimization method proposed as an alternative to particle swarm optimization when direct attraction to a global leader causes the population to lose diversity. Agents are randomly paired each epoch; winners keep their positions, while losers update velocity and position using inertia, attraction to their opponent, and a weighted pull toward the swarm’s center. Positions are clamped to the search bounds.

It describes the single tuning parameter φ and gives dimensionality-based recommendations, then outlines a reproducible MQL5 implementation and tests on Hilly, Forest, and Megacity functions at several dimensions. The stated evaluation criteria are objective-function calls and stability across runs, but the supplied excerpt omits the detailed results, so it does not establish which settings or benchmarks CSO wins. The method is presented for optimization rather than direct trading, and the article notes that its algorithm implementations may differ from canonical versions. It suggests adding personal-best memory as a possible extension.

Key ideas

• CSO randomly pairs agents and updates only the less fit member of each pair.
• Losers learn from their opponent and the population centroid, while winners remain unchanged for that epoch.
• The update combines inertia, pairwise attraction, and a tunable attraction toward the swarm center.
• The article compares solution quality and run-to-run stability on benchmark functions, but the excerpt omits detailed test results.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.