Massive MIMO as an Extreme Learning Machine

This work shows that a massive multiple-input multiple-output (MIMO) system with low-resolution analog-to-digital converters (ADCs) forms natural extreme learning machine (ELM). The receive antennas at the base station serve as hidden nodes of ELM, and ADCs act ELM activation function. By adding random biases to received signals optimizing output weights, can effectively tackle hardware impairments, such nonlinearity power amplifiers ADCs. Moreover, fast adaptive capability allows design an receiver address time-varying effects MIMO channels. Simulations demonstrate promising performance ELM-based compared conventional receivers in dealing impairments.