Federated learning method based on hierarchical tensor decomposition in edge calculation

Abstract

The invention relates to a federated learning method based on hierarchical tensor decomposition in edge calculation. The federated learning method comprises the steps: S1, designing an effective deepneural network sharing model at a cloud end; S2, compressing the designed sharing model according to a hierarchical tensor decomposition method to obtain a hierarchical sharing model; S3, designing aforward propagation algorithm and a reverse propagation algorithm corresponding to the hierarchical sharing model; S4, initializing the hierarchical sharing model at the cloud and issuing the hierarchical sharing model to edge nodes participating in training; S5, learning the hierarchical sharing model obtained in the step S2 by the edge nodes participating in training according to the algorithm designed in the step S3 by using the local data set; and S6, aggregating the edge models at the cloud in an average aggregation mode. According to the federated learning method, distributed training ofthe shared model is realized on the premise of protecting user privacy, and the demand for network bandwidth during distributed training is reduced, and the communication energy consumption of edge nodes is reduced.

Description

technical field [0001] The invention relates to a federated learning method based on hierarchical tensor decomposition in edge computing Background technique [0002] With the rapid development of Internet of Things technology and its wide application in industrial fields such as smart factories, industrial automation, and intelligent manufacturing, Industrial Internet of Things technology has attracted extensive attention from academia and industry. In the Industrial Internet of Things, there is an explosion of data generated by various connected devices. However, it is impractical to directly transmit large amounts of data to remote cloud platforms for further processing and analysis, which may cause severe network congestion and unbearable transmission delays. In recent years, with the rise of edge computing technology, edge devices (nodes) such as sensors and factory gateways have the ability to store, process and analyze local data. In addition, edge devices can colla...

AI Technical Summary

Problems solved by technology

However, training deep learning models on edge devices has the following major disadvantages: On the one hand, due to the growing awareness of data security and user privacy, uploading local datasets on each edge device to cloud servers has Risk of data breach

The second type of method is to reduce the communication bandwidth by quantizing the gradient to a low-precision value through gradient quantization

Images