A Communication-Efficient Adaptive Algorithm for Federated Learning under Cumulative Regret

Sudeep Salgia, Tamir Gabay, Qing Zhao, Kobi Cohen

Research output: Contribution to journalArticlepeer-review

Abstract

We consider the problem of online stochastic optimization in a distributed setting with <italic>M</italic> clients connected through a central server.We develop a distributed online learning algorithm that achieves order-optimal cumulative regret with low communication cost measured in the total number of bits transmitted over the entire learning horizon. This is in contrast to existing studies which focus on the offline measure of simple regret for learning efficiency. The holistic measure for communication cost also departs from the prevailing approach that <italic>separately</italic> tackles the communication frequency and the number of bits in each communication round.

Original languageEnglish
Pages (from-to)1-11
Number of pages11
JournalIEEE Transactions on Signal Processing
DOIs
StateAccepted/In press - 1 Jan 2024

Keywords

  • Communication Efficiency
  • Costs
  • Cumulative Regret
  • Estimation
  • Federated learning
  • Federated learning
  • Noise measurement
  • Optimization
  • Servers
  • Signal processing algorithms

ASJC Scopus subject areas

  • Signal Processing
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'A Communication-Efficient Adaptive Algorithm for Federated Learning under Cumulative Regret'. Together they form a unique fingerprint.

Cite this