A model training method, device and storage medium based on federated learning

Abstract

This application discloses a model training method based on federated learning, which is applied in the field of machine learning. This application includes dividing the K participant nodes into clusters according to the response time of each participant node, and obtaining N clusters; select target clusters from N clusters; select Ω participant nodes from target clusters; send gradient update request to Ω participant nodes, so that Ω participant nodes According to the gradient update request, update the first local model gradient of the local model to obtain the second local model gradient; when receiving the second local model gradient sent by T participant nodes, according to the T second local model gradient Model training on the joint model. The embodiment of the present application provides a device and a storage medium. This application can not only reduce the total waiting time of the server, but also improve the effect of model training.

Description

technical field [0001] The present application relates to the technical field of machine learning, and in particular to a model training method, device and storage medium based on federated learning. Background technique [0002] Federated-learning is a distributed machine learning training method based on privacy protection. The goal of federated learning is to achieve common modeling on the basis of ensuring data privacy security and legality, which can protect data privacy while Train a high-quality centralized artificial intelligence model. [0003] At present, for the efficiency optimization problem of federated learning, the node selection method can be used for optimization. The specific process is that the server selects participant nodes with sufficient equipment resources for joint model training and parameter iteration, and excludes participant nodes with insufficient equipment resources, that is, eliminates participant nodes with slow response time to improve th...

AI Technical Summary

Problems solved by technology

[0004] However, although the elimination of participant nodes with slow response time can improve the training efficiency of federated learning to a certain extent, the sample data of these participant nodes does not participate in the joint model training, so it will lead to deviations in the model training effect , reducing the accuracy of the model prediction

Images