A survey on federated learning

Abstract

Federated learning is a set-up in which multiple clients collaborate to solve machine learning problems, which is under the coordination of a central aggregator. This setting also allows the training data decentralized to ensure the data privacy of each device. Federated learning adheres to two major ideas: local computing and model transmission, which reduces some systematic privacy risks and costs brought by traditional centralized machine learning methods. The original data of the client is stored locally and cannot be exchanged or migrated. With the application of federated learning, each device uses local data for local training, then uploads the model to the server for aggregation, and finally the server sends the model update to the participants to achieve the learning goal. To provide a comprehensive survey and facilitate the potential research of this area, we systematically introduce the existing works of federated learning from five aspects: data partitioning, privacy mechanism, machine learning model, communication architecture and systems heterogeneity. Then, we sort out the current challenges and future research directions of federated learning. Finally, we summarize the characteristics of existing federated learning, and analyze the current practical application of federated learning.