Method and device for multi-party joint training of graph neural network

Abstract

The embodiment of the invention provides a method and device for multi-party joint training of a graph neural network. The multiple parties comprise a plurality of data owners and a server; the graphneural network comprises a graph embedding sub-network and a classification sub-network. And each data owner maintains a part of the graph embedding sub-network, and the server maintains the classification sub-network. Any data owner calculates a primary embedding vector of a sample in a graph embedding sub-network maintained by the data holder through multi-party security calculation MPC, and performs multi-level neighbor aggregation on nodes according to a local graph structure to obtain a high-order embedding vector of the nodes, and sends the high-order embedding vector to a server. And the server synthesizes the high-order embedded vectors from the data owners by using the classification sub-network, and performs classification prediction according to the high-order embedded vectors to determine the loss. And the loss gradient is transmitted back to the graph embedding sub-network in the data holder from the classification sub-network in the server, so that the joint training of the whole graph neural network is realized. According to the embodiment of the invention, the data privacy of each party is protected.

Description

Technical field [0001] One or more embodiments of this specification relate to the fields of data security and machine learning, and in particular, to methods and devices for multi-party joint training of graph neural networks. Background technique [0002] The data needed for machine learning often involves multiple fields. For example, in a user classification analysis scenario based on machine learning, the electronic payment platform owns the user's transaction flow data, the social platform owns the user's friend contact data, and the banking institution owns the user's loan data. Data often exists in the form of islands. Due to industry competition, data security, user privacy, and other issues, data integration is facing great resistance. It is difficult to integrate data scattered on various platforms to train machine learning models. Under the premise of ensuring that data is not leaked, the use of multi-party data to jointly train machine learning models has become a ...

AI Technical Summary

Problems solved by technology

However, this also makes the graph ne

Images