Adapting Federated Learning to Dynamic and Non-Stationary Data: An Online Learning Approach

Online learning-based optimization represents a novel approach that integrates the principles of federated learning and online learning to effectively handle dynamic data distributions and continuous learning scenarios. This paper offers a comprehensive survey of the latest developments in online federated learning, focusing on optimizing communication efficiency in federated learning environments through online learning techniques. We delve into foundational concepts, key methodologies, challenges, and diverse applications of this emerg-ina field.