A density adjustment based particle swarm optimization learning algorithm for neural network design

In this paper, a density adjustment based Particle swarm optimization algorithm is proposed to solve the problem of premature convergence and global optimal in traditional Particle swarm optimization algorithm. Measure the density of particle swarm by entropy, and update the particle swarm to maintain the swarm diversity, which can also help to improve the ability of global optimization. At the same time extend the particle swarm to improve the local optimization capability. Using 1500 remote sensing images including city, mountain and ocean three types of surface feature, compare the training results of neural network classifier trained by BP learning, standard particle swarm optimization and density adjustment based Particle swarm optimization algorithm. The classification results show that the new algorithm converges much faster, and has stronger global optimization ability.