Deep Learning-Aided Multicarrier Systems

This paper proposes a deep learning (DL)-aided multicarrier (MC) system operating on fading channels, where both modulation and demodulation blocks are modeled by neural networks (DNNs), regarded as the encoder decoder of an autoencoder (AE) architecture, respectively. Unlike existing AE-based systems, which incorporate domain knowledge channel equalizer to suppress effects wireless proposed scheme, termed MC-AE, directly feeds with state information received signal, then processed in fully data-driven manner. new approach enables MC-AE jointly learn optimize diversity coding gains over channels. In particular, block error rate is analyzed show its higher performance than hand-crafted baselines, such various recent index modulation-based MC schemes. We extend multiuser scenarios, wherein resultant MU-MC-AE. Accordingly, two novel DNN structures for uplink downlink MU-MC-AE transmissions proposed, along cost function that ensures fast training convergence fairness among users. Finally, simulation results provided superiority DL-based schemes current terms receiver complexity.