Current Best Opposition-Based Learning Salp Swarm Algorithm for Global Numerical Optimization

Abstract

The salp swarm algorithm is one of the novel swarm intelligence metaheuristics. The work proposed in this paper provides further improvements of the salp swarm algorithm, that have been achieved by modifications of the original approach. By analyzing solutions’ quality and convergence speed of basic salp swarm during practical simulations, it was concluded that the exploitation process can be improved. Improvements were achieved by introducing the concept of opposite solutions in the initialization phase, as well as in iterative search process, where a fine-tuned exploitation of the current best solution is performed by generating its opposite individual. Proposed improved salp swarm algorithm was tested on thirteen well-known global benchmarks. Comparative analysis was performed with seven other modern metaheuristics methods, and against the original salp swarm algorithm. Accomplished results have proven that proposed approach in a large degree outscores original algorithm and other approaches included in comparative analysis.