A Deep Neural Network Approach to Max-Min Fair Precoder for Multiple Antenna Systems

This study uses a supervised machine learning-based approach to analyze a low-complexity precoder design to maximize the minimal Signal-to-Interference plus Noise Ratio (SINR) per stream of the multiple-antenna system. The work aims to reduce the amount of complexity in the precoder design as much as possible. To be more specific, the suggested precoder is a two stage technique. The first design's stage makes use of a deep-learning-assisted multi-antenna framework, which is the standard approach, in order to anticipate the uplink power allocation (PA) in the system (based on specific users' service quality (SoQ) targets). While in the second step of the process, an optimization of the downlink Max-min precoder vector is performed using the Lagrangian duality theory. According to the findings of the research, consumers whose expectations regarding the quality of the service are not very high should consider putting into place an adaptive system that uses a deep neural method. When compared to the zero-forcing strategy, our approach results in a performance boost of approximately 2.5 dB when the minimum SINR target is set to 5 dB. For users that are more annoyed with the quality of the service they receive, on the other hand, we have the option of employing a plain zero-forcing technique. In addition, we are able to draw the conclusion that the suggested method introduces a balance between the ease of use and the amount of calculation required,