Extreme Learning Machine-Based Receiver for Multi-User Massive MIMO Systems

Abstract

An extreme learning machine (ELM)-based receiver for multi-user massive MIMO systems is introduced. The proposed ELM combining method, defined in the complex plane, is designed to directly perform MIMO combining processing to the received uplink signals, based on the adoption of the pilot symbols as training data. Numerical results show that by appropriately setting the number of hidden neurons, the ELM achieves higher spectral efficiency and smaller BER, with fewer floating-point operations than the conventional linear MIMO receivers, namely the minimum mean squared error and maximum ratio receivers.