Parameter range estimation using particle swarm optimization

Abstract

Estimation of best range or ranges of the model parameters is achieved by introducting modifications to the Particle Swarm Optimization (PSO) method. A local swarm is introduced by reducing the swarm neighborhood by a finite vector radius R. The global best of the local swarm is then determined from the particles within this neighborhood. We also introduced a random velocity to each particle at every iteration, preventing the swarms from settling prematurely at solutions. The randomness also makes the modified PSO useful for problems where uncertainty in the optimum parameter set or the number of possible parameter sets obtained are relevant. The variance in the solutions found by the PSO algorithm provide an estimate of the parameters’ best range