Design of feed forward neural network based on improved particle swarm optimizer

This paper proposed an improved particle swarm optimizer(PSO) which introduced into hierarchical structure theory and ant colony optimization.The new algorithm updated particles' control gene by releasing pheromone to achieve information sharing between particles.The inertial weight of PSO decreased as exponential curve and added a random disturb in velo-city of the worst particle by far to resolve the premature problem.This new algorithm shows a good performance in optimizing the topology structure and weights of the feedforward neural network synchronously.