A Hybrid Optimization Algorithm for Extreme Learning Machine

Abstract

In this paper, a learning algorithm based on particle swarm optimization method (PSO) and a novel heuristic optimization method of gravitational search algorithm (GSA) for extreme learning machine (ELM) is proposed in terms of improving the generalization performance of single hidden-layer feed-forward neural networks, which is called as PSOGSA-ELM learning algorithm. The proposed learning algorithm uses a hybrid approach of PSO and GSA to select the optimal hidden biases and input weights of ELM, and then the output weights of ELM is analytically determined by the Moore-Penrose generalized inverse. The performance of the proposed algorithm is verified by regression and classification benchmark problems and is compared with PSO–ELM, GSA–ELM, and the original ELM learning algorithms, simulation results show that the proposed algorithm performs equal to or better than the other algorithms in terms of generalization performance and has good convergence speed.