A Survey on Optimization of Federated Learning Based on Edge-Cloud Collaboration

Abstract

In order to address issues including resource limitations, data heterogeneity, and computational inefficiency, this study explores the optimization of lightweight federated learning (FL) in edge-cloud collaborative contexts. Due to non-IID data distributions, high communication costs, and restricted device capabilities, traditional federated learning frameworks perform below expectations in edge environments. This study analyzes the background and challenges, elaborates on definitions and theories, emphasizes the classification of federated learning and finally provides several optimization solutions. The objective is to provide strong support for federated learning applications in resource-constrained scenarios while efficiently increasing the operational efficiency of federated learning systems in edge-cloud environments. The results lay the groundwork for future studies on scalable and effective distributed learning systems by highlighting the promise of lightweight FL in applications like smart healthcare and industrial Internet of Things (IoT). In particular, it underscores the need for adaptive mechanisms that dynamically adjust computation and communication strategies based on the capabilities of edge devices.