Holistic Interference Management for Wireless Networks in the Era of Artificial Intelligence

Future networks are expected to exhibit intense use of artificial intelligence due to the increasing use of intelligent devices in domestic and industrial life. The intelligent devices will communicate with networks and exchange information about expected performance, available cost packages, and availability of network resources along the destination. Therefore, networks need intelligent techniques to learn the state of various network functions and resources and adjust their configurations in an automated way. Machine learning techniques allow the networks to realize such learning and automate the optimization of the network functions and resources. Several techniques have been discussed in the literature to optimize and manage interference in radio networks. However, the existing approaches generally optimize one or a few aspects in a stand-alone fashion. Recently introduced global learning and deep holistic learning techniques can optimize the network function considering all known aspects. This article proposes a novel holistic learning and optimization technique for interference management in wireless networks. It uses a novel objective functions-based feature engineering process to capture the effects of various parameters and actions related to interference management. Transfer learning reduces computational complexity, and ensemble learning aggregates knowledge from base learners corresponding to each objective function. The experimental network is constructed using the NS3 LENA module, and standard Python libraries are used to implement the base learners and proposed model. It uses several base learners that learn the information from possible interference variables and determine the optimal actions across the cells. The experimental results show that the holistic learning-based approach efficiently manages the interference, improves the system capacity, and reduces the interference caused by user arrivals twofold compared to the state-of-the-art techniques.