Communication-Efficient Design for Quantized Decentralized Federated Learning

被引:7
|
作者
Chen, Li [1 ]
Liu, Wei [1 ]
Chen, Yunfei [2 ]
Wang, Weidong [1 ]
机构
[1] Univ Sci & Technol China, Dept Elect Engn & Informat Sci, Hefei 230027, Peoples R China
[2] Univ Durham, Dept Engn, Durham DH1 3LE, England
基金
中国国家自然科学基金;
关键词
Decentralized federated learning; doubly-adaptive quantization; Lloyd-Max quantizer; GRADIENT DESCENT;
D O I
10.1109/TSP.2024.3363887
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
Decentralized federated learning (DFL) is a variant of federated learning, where edge nodes only communicate with their one-hop neighbors to learn the optimal model. However, as information exchange is restricted in a range of one-hop in DFL, inefficient information exchange leads to more communication rounds to reach the targeted training loss. This greatly reduces the communication efficiency. In this paper, we propose a new non-uniform quantization of model parameters to improve DFL convergence. Specifically, we apply the Lloyd-Max algorithm to DFL (LM-DFL) first to minimize the quantization distortion by adjusting the quantization levels adaptively. Convergence guarantee of LM-DFL is established without convex loss assumption. Based on LM-DFL, we then propose a new doubly-adaptive DFL, which jointly considers the ascending number of quantization levels to reduce the amount of communicated information in the training and adapts the quantization levels for non-uniform gradient distributions. Experiment results based on MNIST and CIFAR-10 datasets illustrate the superiority of LM-DFL with the optimal quantized distortion and show that doubly-adaptive DFL can greatly improve communication efficiency.
引用
收藏
页码:1175 / 1188
页数:14
相关论文
共 50 条
  • [1] Communication-efficient and Scalable Decentralized Federated Edge Learning
    Yapp, Austine Zong Han
    Koh, Hong Soo Nicholas
    Lai, Yan Ting
    Kang, Jiawen
    Li, Xuandi
    Ng, Jer Shyuan
    Jiang, Hongchao
    Lim, Wei Yang Bryan
    Xiong, Zehui
    Niyato, Dusit
    PROCEEDINGS OF THE THIRTIETH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, IJCAI 2021, 2021, : 5032 - 5035
  • [2] Communication-efficient Federated Learning via Quantized Clipped SGD
    Jia, Ninghui
    Qu, Zhihao
    Ye, Baoliu
    WIRELESS ALGORITHMS, SYSTEMS, AND APPLICATIONS, WASA 2021, PT I, 2021, 12937 : 559 - 571
  • [3] Communication-Efficient Federated Learning via Quantized Compressed Sensing
    Oh, Yongjeong
    Lee, Namyoon
    Jeon, Yo-Seb
    Poor, H. Vincent
    IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, 2023, 22 (02) : 1087 - 1100
  • [4] Lazily Aggregated Quantized Gradient Innovation for Communication-Efficient Federated Learning
    Sun, Jun
    Chen, Tianyi
    Giannakis, Georgios B.
    Yang, Qinmin
    Yang, Zaiyue
    IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2022, 44 (04) : 2031 - 2044
  • [5] Communication-efficient federated learning
    Chen, Mingzhe
    Shlezinger, Nir
    Poor, H. Vincent
    Eldar, Yonina C.
    Cui, Shuguang
    PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA, 2021, 118 (17)
  • [6] On the Design of Communication-Efficient Federated Learning for Health Monitoring
    Chu, Dong
    Jaafar, Wael
    Yanikomeroglu, Halim
    2022 IEEE GLOBAL COMMUNICATIONS CONFERENCE (GLOBECOM 2022), 2022, : 1128 - 1133
  • [7] A Layer Selection Optimizer for Communication-Efficient Decentralized Federated Deep Learning
    Barbieri, Luca
    Savazzi, Stefano
    Nicoli, Monica
    IEEE ACCESS, 2023, 11 : 22155 - 22173
  • [8] Communication-Efficient Personalized Federated Edge Learning for Decentralized Sensing in ISAC
    Zhu, Yonghui
    Zhang, Ronghui
    Cui, Yuanhao
    Wu, Sheng
    Jiang, Chunxiao
    Jing, Xiaojun
    2023 IEEE INTERNATIONAL CONFERENCE ON COMMUNICATIONS WORKSHOPS, ICC WORKSHOPS, 2023, : 207 - 212
  • [9] FedUVeQCS: Universal Vector Quantized Compressive Sensing for Communication-Efficient Federated Learning
    Liu, Zhengming
    Wang, Hui
    Li, Xiaobo
    IEEE INTERNET OF THINGS JOURNAL, 2024, 11 (22): : 37045 - 37056
  • [10] Communication-Efficient Vertical Federated Learning
    Khan, Afsana
    ten Thij, Marijn
    Wilbik, Anna
    ALGORITHMS, 2022, 15 (08)