Machine Learning based Antenna Selection and Power Allocation in Multi-user MISO Systems

We investigate the performance of multi-user multiple-antenna downlinks via joint antenna selection and power control design. In order to fully exploit the spatial diversity while minimizing the energy consumed by active radio frequency (RF) modules, a subset of antennas are selected to serve the users. Firstly, we propose a joint antenna selection and power allocation (JASPA) algorithm to maximize the system sum rate subjected to the total transmit power constraint and quality of service (QoS) requirements. JASPA copes with the non-convexity of the formulated problem via a doubly iterative algorithm, in which an inner iteration successively optimizes the transmit power followed by an outer loop that tries all valid antenna combinations. Although approaching the global optimality, JASPA suffers a combinatorial complexity, which might limit its application in real-time network operations. To overcome this limitation, we propose a learning-based antenna selection and power allocation (L-ASPA) which significantly reduces the high computational time of JASPA while retaining comparative performance. The core idea behind L-ASPA is to exploit the advances in machine learning to establish underlaying relation between the key system parameters and the selected antennas. The effectiveness of the proposed algorithms is demonstrated via numerical results, which show that JASPA could achieve 90% of the optimal performance while reducing more than 93% computation time.