Machine Learning in 5G/6G Networks: Assessing Deep Neural Network Performance for Sustainable Mobile Communication

The advancement of telecommunication networks, marked by the transition from 5G to 6G technologies, indicates a significant transformation in how we connect, communicate, and spread information. In this thesis we investigate the deep integration of machine learning (ML) and deep neural networks (DNN) within this technological change, outlining their pivotal role in strengthening modern communication networks. The research offers a detailed overview of the strides in 5G/6G technologies and emphasises the critical role of ML in enhancing communication infrastructures to meet escalating data traffic needs. A specific focus is placed on NVIDIA's contributions, mainly through their Sionna library. This tool not only serves as a vital link in connecting 5G Physical layer simulations with advanced ML toolkits but also showcases Nvidia’s commitment to leading innovations in telecommunications. An integral component of this research is the exploration of pruning techniques within neural networks. By removing unnecessary weights, pruning optimises DNN performance without sacrificing accuracy, paving the way for more efficient and agile communication frameworks. Using a thorough methodology, the study compares the effectiveness of pruned DNNs against their unpruned counterparts and simpler models, assessing decision quality, energy efficiency, and scalability.