A parameter adaptive particle swarm optimization algorithm for extreme learning machine

In this paper, a learning algorithm called APSO-ELM for single hidden layer feed-forward neural networks is proposed, in which the input weights and hidden biases are determined by the parameter adaptive particle swarm optimization technique. The performance of the proposed algorithm is verified by simulation of four function approximation and classification benchmark problems. Simulation results show that the proposed algorithm has better global approximation performance and generalization capability. And for a better elucidation of the effectiveness of the proposed algorithm, the algorithm is applied to the robot execution failures problem. Compared with the original ELM algorithm and existing prediction methods of robot execution failures, the classification rate of the proposed algorithm significantly improved and in 10 times´ simulation, the proposed algorithm has nine times 100% classification rate, which promotes the practical application of the algorithm in the field of robotics.