Accelerate Federated Learning of Big Data Analysis Through Distillation and Coupled Gradient Methods

Aiming at the problem that the training time of federated learning (FL) deep neural network is very long or not feasible, this paper presents an efficient federated learning and training framework. The framework consists of two parts: distillation and coupled gradient algorithm, which can reduce the...

Full description

Saved in:
Bibliographic Details
Main Authors: Zhou, Zhiwei, Liu, Weikai, Zhong, Xiaoying
Format: Conference Proceeding
Language:English
Subjects:
Computational modeling
Costs
Coupling Gradient
Deep learning
Distillation
Federated learning
Gradient methods
Neural networks
Training
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Aiming at the problem that the training time of federated learning (FL) deep neural network is very long or not feasible, this paper presents an efficient federated learning and training framework. The framework consists of two parts: distillation and coupled gradient algorithm, which can reduce the computational cost, speed up the training process of federated learning and save computational resources. Simulation experiments were carried out using an open data set(mnist), and the results show that the framework can guarantee the learning performance and control the training cost well.
ISSN:2161-2927
DOI:10.23919/CCC55666.2022.9902345