A Comparative Performance Analysis of Extreme Learning Machine and Echo State Network for Wireless Channel Prediction

In this work, a comparative performance analysis of an extreme learning machine (ELM) and an echo state network (ESN) for forecasting of wireless channel conditions is carried out. These two algorithms are applied to predict signal-to-noise ratio (SNR) for single-input single-output (SISO) system in both picocellular and microcellular environments. Performance indicators used to gain insight into accuracy and effectiveness of ELM and ESN techniques are normalized mean squared error (NMSE) and time consumption. The experimental results performed on measured SNR values show that the ESN algorithm is characterized by shorter test time and higher prediction accuracy in picocellular environment, while the ELM model is recommended for channel prediction in environment which is less frequency selective.