ABSTRACT
The next generation of mobile technology, 6G, will provide users with a ‘hyper-connectivity’ experience in which everything will be interconnected. In the evolution of future hyper-connected networks, green communication for ‘Sustainable Development’ has become a global topic. Revolutionary technologies such as Machine Learning (ML) and Natural Language Processing are at the core of digital transformation, and much of the Green computing is driven by improving ML algorithms. Next-generation communication networks will have limited radio, computing, and energy resources, and the industry is relying on Machine learning to overcome this limitation Beyond 5G networks. High-quality data is needed for a high-performance machine learning model, and high-quality user-generated information is sensitive and hard to collect. This chapter highlighted the effective, energy-efficient, and privacy-protecting distributed learning techniques in communication networks. Current wireless network learnings, which entail centralizing the training data, are inefficient because they require end devices to send their collected data to a central server continuously. Distributed and Federated Learning (FL) enables end devices to effectively train ground-truth data on-device, with only model update parameters sent back to the federated server. This chapter proposes a Chameleon Federated Learning (CFL) framework-based, ‘GREEN6G’ model for network slicing in Beyond 5G systems to solve complex load estimations problems without collecting sensitive private data from end devices. Simulations demonstrated a 61% improvement in Mean Squared Error and a 26% faster convergence with fewer epochs making the proposed GREEN6G model energy-efficient for 5G and beyond systems.
