Author(s)

Heng Zhang

Corresponding Author

Heng Zhang

Abstract

Extreme Learning Machine (ELM) is a new type of feedforward neural network (FNN). Compared with traditional single hidden layer FNN，ELM possesses higher speed and better performance. Due to the random determination of the input weights and hidden layer biases, ELM may need more hidden neurons to achieve a reasonable accuracy. In this paper, a novel ELM learning algorithm optimized by Particle Swarm Optimization (PSO) which called PSO-ELM is proposed. PSO-ELM uses PSO algorithm to select input weights and hidden layer biases. The experimental results based on SinC benchmarking function and a large number of classification experiments from UCI standard data set show that the new algorithm can produce better generalization performance with less hidden neurons than traditional ELM algorithms and conventional feedforward neural networks.

Keywords

Parameters optimizing, Particle swarm optimization, Extreme learning machine, Hidden neurons.