关键词: 联邦学习, 深度学习, 通信效率, 隐私保护, 机器学习

Abstract: Abstract: Federated learning is a distributed machine learning framework that effectively addresses the data silo problem and is crucial for ensuring privacy protection for individuals and organizations. However, enhancing the efficiency of federated learning remains a pressing issue due to its inherently high costs. This paper provides a comprehensive summary and investigation of current mainstream research on improving the efficiency of federated learning. It reviews the background of efficient federated learning, including its origins and core concepts, and explains the classifications within federated learning. The paper delves into the efficiency challenges in federated learning, categorizing them into heterogeneous problems, personalized problems, and communication cost issues. Furthermore, this paper analyzes and discusses detailed solutions to these efficiency problems, categorizing the research into two main areas: model compression optimization methods and communication optimization methods. Additionally, it outlines the existing challenges in achieving efficient federated learning and presents future research directions in this field. By addressing these issues, this study aims to provide a roadmap for advancing the efficiency and practicality of federated learning in various applications, ultimately making it a more viable solution for privacy-preserving machine learning.

Key words: federated learning, deep learning, communication efficiency, privacy protection, machine learning