Variance Reduced Diffusion Adaptation for Online Learning over Networks

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

2 Scopus citations

Abstract

The stochastic variance reduced gradient (SVRG) algorithm has shown its effectiveness in accelerating the convergence of stochastic gradient algorithms. Considering the emergent applications of distributed estimation, it is interesting to investigate the way to adapt this algorithm to distributed learning with streaming data. For this purpose, in this work we first propose a time-averaging SVRG algorithm that fits into the context of streaming data processing. Then, we integrate this algorithm with the diffusion adaptation to enhance the performance of distributed estimation over networks. Theoretical analysis of the resulted algorithm is conducted to characterize its stability. We also provide the simulation results to illustrate its favorable performance.

Original languageEnglish
Title of host publicationICSPCC 2020 - IEEE International Conference on Signal Processing, Communications and Computing, Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781728172019
DOIs
StatePublished - 21 Aug 2020
Event2020 IEEE International Conference on Signal Processing, Communications and Computing, ICSPCC 2020 - Macau, China
Duration: 21 Aug 202023 Aug 2020

Publication series

NameICSPCC 2020 - IEEE International Conference on Signal Processing, Communications and Computing, Proceedings

Conference

Conference2020 IEEE International Conference on Signal Processing, Communications and Computing, ICSPCC 2020
Country/TerritoryChina
CityMacau
Period21/08/2023/08/20

Keywords

  • diffusion strategy
  • Distributed estimation
  • stochastic optimization
  • SVRG
  • variance reduction

Fingerprint

Dive into the research topics of 'Variance Reduced Diffusion Adaptation for Online Learning over Networks'. Together they form a unique fingerprint.

Cite this