Abstract: Federated learning is a distributed training framework that can build a collection of decentralized data and combine multi-party data, and then use the encrypted intermediate parameters to complete the joint training of multiple data sources for the joint model. Based on the practice of Huawei NAIE Federated Learning Service, this paper discusses the key technologies and typical applications of the implementation of the Federated Learning, and points out the development direction of Federated Learning on the basis of the practice in the telecommunication field.

Key words: federated learning, privacy protection, model localization