Rate distortion optimization for adaptive gradient quantization in federated learning

Federated Learning (FL) is an emerging machine learning framework designed to preserve privacy. However, the continuous updating of model parameters over uplink channels with limited throughput leads to a huge communication overload, which is a major challenge for FL. To address this issue, we propo...

Full description

Saved in:
Bibliographic Details
Main Authors: Guojun Chen, Kaixuan Xie, Wenqiang Luo, Yinfei Xu, Lun Xin, Tiecheng Song, Jing Hu
Format: Article
Language:English
Published: KeAi Communications Co., Ltd. 2024-12-01
Series:Digital Communications and Networks
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S235286482400018X
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846101649186095104
author Guojun Chen
Kaixuan Xie
Wenqiang Luo
Yinfei Xu
Lun Xin
Tiecheng Song
Jing Hu
author_facet Guojun Chen
Kaixuan Xie
Wenqiang Luo
Yinfei Xu
Lun Xin
Tiecheng Song
Jing Hu
author_sort Guojun Chen
collection DOAJ
description Federated Learning (FL) is an emerging machine learning framework designed to preserve privacy. However, the continuous updating of model parameters over uplink channels with limited throughput leads to a huge communication overload, which is a major challenge for FL. To address this issue, we propose an adaptive gradient quantization approach that enhances communication efficiency. Aiming to minimize the total communication costs, we consider both the correlation of gradients between local clients and the correlation of gradients between communication rounds, namely, in the time and space dimensions. The compression strategy is based on rate distortion theory, which allows us to find an optimal quantization strategy for the gradients. To further reduce the computational complexity, we introduce the Kalman filter into the proposed approach. Finally, numerical results demonstrate the effectiveness and robustness of the proposed rate-distortion optimization adaptive gradient quantization approach in significantly reducing the communication costs when compared to other quantization methods.
format Article
id doaj-art-aa5f073d876b478a99b5c0c69d4825fc
institution Kabale University
issn 2352-8648
language English
publishDate 2024-12-01
publisher KeAi Communications Co., Ltd.
record_format Article
series Digital Communications and Networks
spelling doaj-art-aa5f073d876b478a99b5c0c69d4825fc2024-12-29T04:47:37ZengKeAi Communications Co., Ltd.Digital Communications and Networks2352-86482024-12-0110618131825Rate distortion optimization for adaptive gradient quantization in federated learningGuojun Chen0Kaixuan Xie1Wenqiang Luo2Yinfei Xu3Lun Xin4Tiecheng Song5Jing Hu6National Mobile Communication Research Laboratory, Southeast University, Nanjing 210096, China; School of Information Science and Engineering, Southeast University, Nanjing, 210096, ChinaChina Mobile Research Institute, Beijing 100053, ChinaSchool of Information Science and Engineering, Southeast University, Nanjing, 210096, ChinaSchool of Information Science and Engineering, Southeast University, Nanjing, 210096, ChinaChina Mobile Research Institute, Beijing 100053, ChinaNational Mobile Communication Research Laboratory, Southeast University, Nanjing 210096, China; School of Information Science and Engineering, Southeast University, Nanjing, 210096, China; Corresponding author at: National Mobile Communication Research Laboratory, Southeast University, Nanjing 210096, China.National Mobile Communication Research Laboratory, Southeast University, Nanjing 210096, China; School of Information Science and Engineering, Southeast University, Nanjing, 210096, ChinaFederated Learning (FL) is an emerging machine learning framework designed to preserve privacy. However, the continuous updating of model parameters over uplink channels with limited throughput leads to a huge communication overload, which is a major challenge for FL. To address this issue, we propose an adaptive gradient quantization approach that enhances communication efficiency. Aiming to minimize the total communication costs, we consider both the correlation of gradients between local clients and the correlation of gradients between communication rounds, namely, in the time and space dimensions. The compression strategy is based on rate distortion theory, which allows us to find an optimal quantization strategy for the gradients. To further reduce the computational complexity, we introduce the Kalman filter into the proposed approach. Finally, numerical results demonstrate the effectiveness and robustness of the proposed rate-distortion optimization adaptive gradient quantization approach in significantly reducing the communication costs when compared to other quantization methods.http://www.sciencedirect.com/science/article/pii/S235286482400018XFederated learningCommunication efficiencyAdaptive quantizationRate distortion
spellingShingle Guojun Chen
Kaixuan Xie
Wenqiang Luo
Yinfei Xu
Lun Xin
Tiecheng Song
Jing Hu
Rate distortion optimization for adaptive gradient quantization in federated learning
Digital Communications and Networks
Federated learning
Communication efficiency
Adaptive quantization
Rate distortion
title Rate distortion optimization for adaptive gradient quantization in federated learning
title_full Rate distortion optimization for adaptive gradient quantization in federated learning
title_fullStr Rate distortion optimization for adaptive gradient quantization in federated learning
title_full_unstemmed Rate distortion optimization for adaptive gradient quantization in federated learning
title_short Rate distortion optimization for adaptive gradient quantization in federated learning
title_sort rate distortion optimization for adaptive gradient quantization in federated learning
topic Federated learning
Communication efficiency
Adaptive quantization
Rate distortion
url http://www.sciencedirect.com/science/article/pii/S235286482400018X
work_keys_str_mv AT guojunchen ratedistortionoptimizationforadaptivegradientquantizationinfederatedlearning
AT kaixuanxie ratedistortionoptimizationforadaptivegradientquantizationinfederatedlearning
AT wenqiangluo ratedistortionoptimizationforadaptivegradientquantizationinfederatedlearning
AT yinfeixu ratedistortionoptimizationforadaptivegradientquantizationinfederatedlearning
AT lunxin ratedistortionoptimizationforadaptivegradientquantizationinfederatedlearning
AT tiechengsong ratedistortionoptimizationforadaptivegradientquantizationinfederatedlearning
AT jinghu ratedistortionoptimizationforadaptivegradientquantizationinfederatedlearning