Deep Phase-Transmittance RBF Neural Network for Beamforming With Multiple Users

In recent years, beamforming has been vital to increasing the spectral and energy efficiency of the current and next-generation wireless communication systems, such as 5G, the Internet of Things (IoT), and beyond. In such a context, this letter extends the previously proposed single user, multiple-input single-output (SU-MISO) phase-transmittance radial basis function (PT-RBF) beamforming to self-organizing wireless network receivers with multiple users, multiple-input multiple-output (MU-MIMO) beamforming. In the proposed novel approach, the PT-RBF is designed to support multiple outputs and multiple layers, an innovation compared to previous shallow single output PT-RBF. On account of this deep neural network architecture, the proposed deep PT-RBF can handle multiple users with different modulation formats and distinct orders, presenting better results when compared either with other complex-valued neural networks or with the normalized least mean square algorithm.