QoS-Aware Power Management with Deep Learning

Network densification is becoming an overwhelming phenomenon in many emerging wireless communication paradigms. Although network densification may promote system metrics like the throughputs, the quality-of-service (QoS) issue needs to be carefully investigated. Commonly, the QoS-aware power management is tightly restricted by the complicated patterns of interferences among multiple active communication devices. Conventional approaches in optimizing the QoS-aware power management problem may either fail to convergence or the overall power rate is unsatisfactory. In this paper, we make an effort to solve the QoS-aware power management problem with the aid of the deep learning (DL) methodology. Recently DL has shone light on wide variety of research fields, such as image processing and natural language processing. It is our intensive interest in exploring the role that DL plays to solve the QoS-aware power management problem. In the presented extensive experimental analysis work, we show that the DL-based method can well match the solution generated by a conventional optimization procedure. It is impressed that the convergence of DL is quite fast. Moreover, the DL-based approach demonstrates the better performance when the conventional method enters the infeasible region.