Power allocation scheme based on support vector machine forDAS and CAS

Abstract

In this paper, we particularly focus on the great potential of applying machine learning algorithms to the power allocation of wireless communications. The upcoming fifth generation (5G) wireless networks forces people to look for new approaches to reduce the occupation of computing resources in communication networks to meet the rapidly growing rate of wireless data while guaranteeing reliable communication. In this paper, by interpreting the power allocation to multi-class classification learning, we develop a power allocation scheme based on support vector machine (SVM) algorithm for the co-located antenna systems (CAS) and the distributed antenna systems (DAS), respectively. We compare the SE performance between the SVM algorithm applied to power allocation and the conventional power allocation algorithm. Simulation results show that the multi-class SVM classifier can obtain the power allocation scheme that is very close to the conventional method (sub-gradient method and bisection method) both in DAS and CAS, and also provides a low-complexity solution to solve the power allocation problem in DAS and CAS, which achieves the trade-off between communication performance and computational complexity.