Federation learning method, device and system based on multiple labels

Abstract

The embodiment of the invention provides a federated learning method, device and system based on multiple labels. The method comprises the following steps: when multiple mechanisms perform federated learning, a trusted execution environment can obtain multiple label data sets provided by the multiple mechanisms, any label data set comprises original labels of multiple users, and at least one userin the multiple users has inconsistent multiple original labels in the multiple label data sets; after a plurality of label data sets are obtained, a preset weak supervised learning algorithm is utilized to carry out learning training on the plurality of label data sets to obtain a unified target label data set, and the target label data set comprises target labels of the plurality of users; and the target label data set is sent to the plurality of institutions, so that the plurality of institutions perform federated learning based on the target label data set.

Description

technical field [0001] This document relates to the field of federated learning, in particular to a multi-label-based federated learning method, device and system. Background technique [0002] Federated learning (also known as federated learning, federated learning) is a machine learning framework that can effectively help multiple organizations perform data usage and machine learning modeling while meeting user privacy protection and data security requirements. [0003] Usually, multiple institutions can perform learning and training based on sample label data and feature data when performing federated learning, where the sample label data can be provided by multiple institutions. However, due to the business scenarios of multiple institutions or different definitions of labels, the label data provided by multiple institutions are usually inconsistent. In this way, when performing federated learning, it will not be possible to determine which institution's label data shall...

AI Technical Summary

Problems solved by technology

[0004] The embodiment of this specification provides a multi-label-based federated learning method, device and system, which is used to solve the problem that in federated learning, when the label data provided by multiple institutions is inconsistent, it is impossible to determine which institution's label data is used for learning. training problem

Images