Optimization of the Particle Swarm Algorithm

摘要

Particle Swarm Optimization is a swarm intelligence based and stochastic algorithm to solve the optimization problem. This paper presents the Multidimensional Particle Swarm algorithm with non-equidistant discrete input data such as E-Series or Renard numbers for circuit design. The authors describe the optimization of this method for different circuit designs, agent recycling, and omission of already computed points. The problem of omission of already computed points is to determinate when it is faster to omit the points than compute them. The personal omission history can be used for agent recycling or trajectory corrections. There is also described effect of recycled agents with corrected parameters on the convergence of optimalization. Parameters corrections are based on the principles of genetic algorithms in the other words inheritance from the best rated agents. System of agents rating is described briefly.