Channel-Driven Monte Carlo Sampling for Bayesian Distributed Learning in Wireless Data Centers

被引:0
|
作者
Liu, Dongzhu [1 ]
Simeone, Osvaldo [1 ]
机构
[1] Kings Coll London, Kings Commun Learning & Informat Proc KCLIP Lab, Dept Engn, London WC2R 2LS, England
基金
欧洲研究理事会;
关键词
Wireless communication; Servers; Bayes methods; Protocols; Data centers; Monte Carlo methods; Distance learning; Distributed Bayesian learning; consensus Monte Carlo; over-the-air computation; federated learning; uncoded transmission; wireless data centers; EDGE; COMMUNICATION; AGGREGATION; DESIGN;
D O I
10.1109/JSAC.2021.3118406
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
Conventional frequentist learning, as assumed by existing federated learning protocols, is limited in its ability to quantify uncertainty, incorporate prior knowledge, guide active learning, and enable continual learning. Bayesian learning provides a principled approach to address all these limitations, at the cost of an increase in computational complexity. This paper studies distributed Bayesian learning in a wireless data center setting encompassing a central server and multiple distributed workers. Prior work on wireless distributed learning has focused exclusively on frequentist learning, and has introduced the idea of leveraging uncoded transmission to enable "over-the-air" computing. Unlike frequentist learning, Bayesian learning aims at evaluating approximations or samples from a global posterior distribution in the model parameter space. This work investigates for the first time the design of distributed one-shot, or "embarrassingly parallel", Bayesian learning protocols in wireless data centers via consensus Monte Carlo (CMC). Uncoded transmission is introduced not only as a way to implement "over-the-air" computing, but also as a mechanism to deploy channel-driven MC sampling: Rather than treating channel noise as a nuisance to be mitigated, channel-driven sampling utilizes channel noise as an integral part of the MC sampling process. A simple wireless CMC scheme is first proposed that is asymptotically optimal under Gaussian local posteriors. Then, for arbitrary local posteriors, a variational optimization strategy is introduced. Simulation results demonstrate that, if properly accounted for, channel noise can indeed contribute to MC sampling and does not necessarily decrease the accuracy level.
引用
收藏
页码:562 / 577
页数:16
相关论文
共 50 条
  • [1] Wireless Federated Langevin Monte Carlo: Repurposing Channel Noise for Bayesian Sampling and Privacy
    Liu, Dongzhu
    Simeone, Osvaldo
    [J]. IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, 2023, 22 (05) : 2946 - 2961
  • [2] Robust Distributed Bayesian Learning with Stragglers via Consensus Monte Carlo
    Chittoor, Hari Hara Suthan
    Simeone, Osvaldo
    [J]. 2022 IEEE GLOBAL COMMUNICATIONS CONFERENCE (GLOBECOM 2022), 2022, : 609 - 614
  • [3] Adaptive Bayesian Sampling with Monte Carlo EM
    Roychowdhury, Anirban
    Parthasarathy, Srinivasan
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 30 (NIPS 2017), 2017, 30
  • [4] Exchange Monte Carlo sampling from Bayesian posterior for singular learning machines
    Nagata, Kenji
    Watanabe, Sumio
    [J]. IEEE TRANSACTIONS ON NEURAL NETWORKS, 2008, 19 (07): : 1253 - 1266
  • [5] Distributed evolutionary Monte Carlo for Bayesian computing
    Hu, Bo
    Tsui, Kam-Wah
    [J]. COMPUTATIONAL STATISTICS & DATA ANALYSIS, 2010, 54 (03) : 688 - 697
  • [6] Bayesian inference, Monte Carlo sampling and operational risk
    Peters, G. W.
    Sisson, S. A.
    [J]. JOURNAL OF OPERATIONAL RISK, 2006, 1 (03): : 27 - 50
  • [7] On sequential Monte Carlo sampling methods for Bayesian filtering
    Arnaud Doucet
    Simon Godsill
    Christophe Andrieu
    [J]. Statistics and Computing, 2000, 10 : 197 - 208
  • [8] SAMPLING SIZE IN MONTE CARLO BAYESIAN COMPRESSIVE SENSING
    Kyriakides, Ioannis
    Pribic, Radmila
    [J]. 2014 IEEE 8TH SENSOR ARRAY AND MULTICHANNEL SIGNAL PROCESSING WORKSHOP (SAM), 2014, : 397 - 400
  • [9] On sequential Monte Carlo sampling methods for Bayesian filtering
    Doucet, A
    Godsill, S
    Andrieu, C
    [J]. STATISTICS AND COMPUTING, 2000, 10 (03) : 197 - 208
  • [10] Bayesian inference of multi-group nuclear data by Monte Carlo sampling method
    Wu, Qu
    Peng, Xingjie
    Yu, Yingrui
    Li, Qing
    [J]. ANNALS OF NUCLEAR ENERGY, 2021, 161