FEDERATED LEARNING METHOD AND APPARATUS FUSING PUBLIC DOMAIN DATA AND PRIVATE DATA, AND SYSTEM

摘要

A federated learning method and apparatus fusing public domain data and private data, and a system. Said method comprises: after integrating model parameters sent by at least two first member nodes (62), a second member node (61) in a federation further updating, on the basis of public domain data, first target model parameters obtained by means of integration, so as obtain second target model parameters, and then sending the second target model parameters to the first member nodes (62) in the federation, so that the first member nodes (62) in the federation perform local training again on the basis of the second target model parameters and local data.