Detection Algorithm Based on Deep Learning for the Multi-user MIMO-NOMA System

Traditional non-orthogonal multiple access (NOMA) applies the Successive Interference Cancellation (SIC) approach to decode the signal superimposed by different users, which is over dependent on the precise of the parameter estimation. To overcome this shortcoming, a new receiver structure for multi-user MIMO-NOMA systems based on deep learning (DL) is proposed, which can restore the transmitted signal from the received signal. In order to speed the convergence of the DL, one new hybrid algorithm and structure of DL and zero-forcing (ZF)/ match-filter (MF) is proposed. The simulation results represent that the simple ZF/MF algorithm can speed the convergence and improve the system performance on the basis of original neural networks.