An analysis of particle swarm optimizers

摘要: Many scientific, engineering and economic problems involve the optimisation of a set parameters. These include examples like minimising losses in power grid by finding optimal configuration components, or training neural network to recognise images people's faces. Numerous algorithms have been proposed solve these problems, with varying degrees success. The Particle Swarm Optimiser (PSO) is relatively new technique that has empirically shown perform well on many problems. This thesis presents theoretical model can be used describe long-term behaviour algorithm. An enhanced version constructed guaranteed convergence local minima. algorithm extended further, resulting an global A for constructing cooperative PSO developed, introduction two PSO-based algorithms. Empirical results are presented support properties predicted various models, using synthetic benchmark functions investigate specific properties. then applied task networks, corroborating obtained functions.