Elite Directed Particle Swarm Optimization with Historical Information for High-Dimensional Problems

High-dimensional optimization problems are ubiquitous in every field nowadays, which seriously challenge the ability of existing optimizers. To solve this kind effectively, paper proposes an elite-directed particle swarm (EDPSO) with historical information to explore and exploit high-dimensional solution space efficiently. Specifically, EDPSO, is first separated into two exclusive sets based on Pareto principle (80-20 rule), namely elite set containing top best 20% particles non-elite consisting remaining 80% particles. Then, further layers same size from worst. As a result, divided three layers. Subsequently, third layer learn those layers, while second layer, condition that remain unchanged. In way, learning effectiveness diversity could be largely promoted. enhance particles, we maintain additional archive store obsolete elites, use predominant elites along direct update layer. With these mechanisms, proposed EDPSO expected compromise search intensification diversification well at level level, space. Extensive experiments conducted widely used CEC’2010 CEC’2013 benchmark problem validate EDPSO. Compared several state-of-the-art large-scale algorithms, demonstrated achieve highly competitive or even much better performance tackling problems.