Alternating Graph-Regularized Neural Network for Adaptive Beamforming in 5G milli meter wave Massive MIMO Multicellular Networks

Abstract

Abstract The idea of ultra-dense networking (UDN) has been more and more popular in recent years as a potential way to increase energy efficiency (EE), spectral efficiency (SE) over tiny geographic areas. However, since channel state information (CSI) is necessary, deployment of large number of Access Points may not only significantly increase signaling burden but also take advantage of the pilot contamination effect. To overcome this issue, present an Alternating Graph-Regularized Neural Network for Adaptive Beamforming in 5G Milli Metre Wave Massive MIMO Multicellular Networks (AGNN-ABM-MIMO-MCN) is proposed. When each active base stations of evaluated orientations use huge multiple-input, multiple-output setups, here, beamforming is carried out with aid of preset set of configurations by appropriately producing higher directional beams on demand, can handle variety of traffic circumstances. Then, Beam forming approach depend on AGNN, which is trained to generate appropriate beam forming configuration. In general, AGNN classifier does not describe modifying optimization techniques to identify optimal parameters to beamforming configuration. Therefore, it is proposed to use Tyrannosaurus Optimization Algorithm (TOA) to optimize Alternating Graph-regularized Neural Network, which accurately generates the beamforming configuration. The proposed method is analyzed with different metrics likes Energy Efficiency, Spectral efficiency, Normalized Mean-Squared Error, Latency, Bit Error Rate, blocking probability and computational complexity. The simulation outcomes proves that the proposed technique attains provides 21.25%, 23.19% and 22.14% lower NMSE, 23.12%, 24.43% and 21.32% lower latency, 23.25%, 22.19% and 25.32% higher Spectral Efficiency while analyzed with existing techniques likes machine learning adaptive beamforming framework for 5G millimeter wave massive MIMO multi cellular networks (MLAB-MMWM-MIMO), deep learning-enabled relay node placement with selection framework in multicellular networks (DL-ERN-MCN) and deep learning framework for beam selection with power control in massive MIMO-millimeter-wave communications (DLF-BSPC-MIMO) respectively.