Machine learning integration in Communication system for efficient selection of signals

Through combined antenna selecting and power management design, we examine the effectiveness of multi-user multiple-antenna downlinks. A fraction of antennas are chosen to assist the customers in order to fully leverage the geographical diversity and reduce the energy spent by operational radio frequency (RF) modules. First, we suggest the joint antenna selection and power allocation (JASPA) method to increase the network sum rate while taking quality of service (QoS) needs and the total transmission power restriction into consideration. JASPA uses a doubly repetitive technique to overcome the non-convexity of the defined issue. In the internal repetition, the transmission power is incrementally optimized, while in the outside loop, all possible antenna configurations are tested. JASPA has a combinatorial complexity issue while being close to the worldwide optimality, which might restrict its usage in real time system operations. We suggest a learning-oriented approach to antenna selection and power allocation (L-ASPA) to get over this restriction while maintaining JASPA's competitive performance. The fundamental concept behind L-ASPA is to take use of recent developments in ML to create a subordinate relationship between the important system variables and the chosen antennas. Numerical findings that show that JASPA could reach 90% of the ideal performance while lowering more than 93% of computing time serve as evidence of the usefulness of the suggested methods.