Lightweight Federated Learning for Energy-Efficient English Corpus Distribution and Optimization in Edge-Cloud Collaboration Networks

This study introduces an energy-aware collaborative architecture that synergistically converges edge computing resources, cloud infrastructure, and privacy-preserving distributed learning mechanisms for optimized English corpus distribution. The proposed framework systematically implements lightweight federated learning through a three-tier optimization paradigm. In particular, by combining federated learning with edge-cloud architecture, we can aggregate the edge information easily and execute the federated learning model naturally, so as to improve performance. In addition, various strategies have been introduced to lightweight the model from the aspects of devices, structure, and quantization. Hence, the lightweight feature is naturally supported in this proposed framework. The proposed framework and method are implemented and tested via comprehensive experiments. The corresponding results indicate we have achieved great performance, including an 83% reduction in energy consumption and a 76% reduction in latency, which state that the proposed method outperforms the state-of-the-art methods.