A federated learning method and system based on parameter replacement algorithm

Abstract

The present invention provides a federated learning method and system based on a parameter replacement algorithm, wherein the method includes: the aggregation server sends initial parameters to participants (edge ​​computing devices and terminal devices), and the participants initialize local models according to the initial parameters; The participants train the local model according to their respective training sets to obtain the local model parameters; the edge computing device sends the trained model parameters to the server and the terminal device; the terminal device initializes the second model according to the updated first model parameters, executes Status judgment, according to the current status, implement different upload strategies to upload parameters; the aggregation server integrates the received local model parameters to obtain the starting parameters of the next round; repeat the above until all second devices exit federated learning. The invention ensures that the parameters uploaded by users in federated learning are visible, avoids malicious users from uploading wrong parameters, and can ensure privacy security and precision security in the federated learning training process.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to a federated learning method and system based on a parameter replacement algorithm. Background technique [0002] Federated Learning (Federated Learning) is a feasible method for training models based on distributed data. This method keeps private data in edge nodes and trains models by sharing parameters, thus preventing the privacy leakage of original data. [0003] The core of federated learning lies in two points: one is to protect the privacy of participants; the other is to effectively learn from the data owned by participants. At present, in federated learning, there is a problem of privacy leakage due to the parameters uploaded by users containing private information, and the idea of ​​the current mainstream solution includes adding a certain amount of noise to the uploaded parameters and encrypting the uploaded parameters. However, the former has poor pri...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a federated learning method and system based on a parameter replacement algorithm to solve the technical problem of precision loss caused by invisible parameters and enhanced parameter noise in traditional privacy protection methods

Images