Extreme learning machine tuning by original sine cosine algorithm

Abstract

Extreme learning machine (ELM) is a revolutionary approach for training single-hidden layer feedforward neural networks that combines both high performance and rapid learning speed. Because the input weights and hidden neurons biases are randomly initialized and stay fixed during the process of learning, and the output weights are analytically calculated. ELM produces high generalization capability with a huge number of hidden neurons. The sine cosine method was presented in this study for tuning the input weights and hidden biases. The suggested method is named SCA-ELM, and it selects the input weights and hidden biases using SCA while determining the output weights using the Moore-Penrose (MP) generalized inverse, The aim is to improve the original extreme learning machine algorithm.The suggested methodologies were evaluated on several benchmark classification data sets, and compared with other recent state-of-art algorithms. Simulations reveal that the suggested method outperforms the other alternatives in the comparative analysis.