Federated learningFederated learningFederated learning is a framework that enables multiple parties to collaboratively train AIArtificial intelligence (AI) models, while each party retains control of its own raw data, never sharing it with others. As such, it resolves some of the legal and technical challenges with basic multiparty computationMultiparty computation. Federated learningFederated learning relies on edge computingEdge computing, a distributed computingDistributed computing model in which processing of raw data takes place on local devices, close to or where it was generated. Only results from local analysis using edge computingEdge computing are sent to external data centers to collaboratively train a global AIArtificial intelligence (AI) model. Variations of the federated learningFederated learning framework are used in a wide range of scenarios, such as cross-device federated learningFederated learning on mobile devices by Internet companies to gain a better understanding of customer behaviors and preferences; cross-silo federated learningFederated learning, cross-siloCross-silo federated learningFederated learning by different institutions with similar data that collaborate to increase the reliability of rare event identification (e.g., financial fraud, identification of diseases); vertical federated learningFederated learningVertical federated learningFederated learning, vertical for aggregating information from different institutions with data about the same customers but different customer attributes, to gain a holistic understanding of customers; and federated transfer learningFederated transfer learningFederated learningTransfer learning to increase the data sample and data attribute sizes by transferring knowledge across data sets.