Hybrid Comprehensive Learning Particle Swarm Optimizer with Adaptive Starting Local Search

Particle Swarm Optimization (PSO) offers efficient simultaneous global and local searches but is challenged with the problem of slow local convergence. To address this issue, a hybrid comprehensive learning PSO algorithm with adaptive starting local search (ALS-HCLPSO) is proposed. Determining when to start local search is the main of ALS-HCLPSO. A quasi-entropy index is innovatively utilized as the criterion of population diversity to depict an aggregation degree of particles and to ascertain whether the global optimum basin has been explored. This adaptive strategy ensures the proper starting of local search. The test results on eight multimodal benchmark functions demonstrate the performance superiority of ALS-HCLPSO. And comparison results on six advanced PSO variants further test the validity and superiority of ALS-HCLPSO algorithm.