Device Sampling for Heterogeneous Federated Learning: Theory, Algorithms, and Implementation

The conventional federated learning (FedL) architecture distributes machine learning (ML) across worker devices by having them train local models that are periodically aggregated by a server. FedL ignores two important characteristics of contemporary wireless networks, however: (i) the network may c...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Annual Joint Conference of the IEEE Computer and Communications Societies s. 1 - 10
Hlavní autoři: Wang, Su, Lee, Mengyuan, Hosseinalipour, Seyyedali, Morabito, Roberto, Chiang, Mung, Brinton, Christopher G.
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 10.05.2021
Témata:
ISSN:2641-9874
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Abstract The conventional federated learning (FedL) architecture distributes machine learning (ML) across worker devices by having them train local models that are periodically aggregated by a server. FedL ignores two important characteristics of contemporary wireless networks, however: (i) the network may contain heterogeneous communication/computation resources, while (ii) there may be significant overlaps in devices' local data distributions. In this work, we develop a novel optimization methodology that jointly accounts for these factors via intelligent device sampling complemented by device-to-device (D2D) offloading. Our optimization aims to select the best combination of sampled nodes and data offloading configuration to maximize FedL training accuracy subject to realistic constraints on the network topology and device capabilities. Theoretical analysis of the D2D offloading subproblem leads to new FedL convergence bounds and an efficient sequential convex optimizer. Using this result, we develop a sampling methodology based on graph convolutional networks (GCNs) which learns the relationship between network attributes, sampled nodes, and resulting offloading that maximizes FedL accuracy. Through evaluation on real-world datasets and network measurements from our IoT testbed, we find that our methodology while sampling less than 5% of all devices outperforms conventional FedL substantially both in terms of trained model accuracy and required resource utilization.
AbstractList The conventional federated learning (FedL) architecture distributes machine learning (ML) across worker devices by having them train local models that are periodically aggregated by a server. FedL ignores two important characteristics of contemporary wireless networks, however: (i) the network may contain heterogeneous communication/computation resources, while (ii) there may be significant overlaps in devices' local data distributions. In this work, we develop a novel optimization methodology that jointly accounts for these factors via intelligent device sampling complemented by device-to-device (D2D) offloading. Our optimization aims to select the best combination of sampled nodes and data offloading configuration to maximize FedL training accuracy subject to realistic constraints on the network topology and device capabilities. Theoretical analysis of the D2D offloading subproblem leads to new FedL convergence bounds and an efficient sequential convex optimizer. Using this result, we develop a sampling methodology based on graph convolutional networks (GCNs) which learns the relationship between network attributes, sampled nodes, and resulting offloading that maximizes FedL accuracy. Through evaluation on real-world datasets and network measurements from our IoT testbed, we find that our methodology while sampling less than 5% of all devices outperforms conventional FedL substantially both in terms of trained model accuracy and required resource utilization.
Author Hosseinalipour, Seyyedali
Morabito, Roberto
Brinton, Christopher G.
Wang, Su
Lee, Mengyuan
Chiang, Mung
Author_xml – sequence: 1
  givenname: Su
  surname: Wang
  fullname: Wang, Su
  email: wang2506@purdue.edu
  organization: Purdue University,School of Electrical and Computer Engineering
– sequence: 2
  givenname: Mengyuan
  surname: Lee
  fullname: Lee, Mengyuan
  email: mengyuan_lee@zju.edu.cn
  organization: Zhejiang University,College of Information Science and Electronic Engineering
– sequence: 3
  givenname: Seyyedali
  surname: Hosseinalipour
  fullname: Hosseinalipour, Seyyedali
  email: hosseina@purdue.edu
  organization: Purdue University,School of Electrical and Computer Engineering
– sequence: 4
  givenname: Roberto
  surname: Morabito
  fullname: Morabito, Roberto
  email: roberto.morabito@princeton.edu
  organization: Princeton University, and Ericsson Research,Department of Electrical Engineering
– sequence: 5
  givenname: Mung
  surname: Chiang
  fullname: Chiang, Mung
  email: chiang@purdue.edu
  organization: Purdue University,School of Electrical and Computer Engineering
– sequence: 6
  givenname: Christopher G.
  surname: Brinton
  fullname: Brinton, Christopher G.
  email: cgb@purdue.edu
  organization: Purdue University,School of Electrical and Computer Engineering
BookMark eNotkMtOwkAYRkejiYA8gZtxT3FuzMUdASskaBfizoRMO39hDJ0h02rC29tE8i3O5uQsviG6CTEAQo-UTCkl5mn9nheL4k0wo-mUEUanRmhtiLxCY6M0lXImCGczdo0GTAqaGa3EHRq27TchRCsmB-hrCb--Avxhm9PRhz2uY8Ir6CDFPQSIPy3OwUGyHTi8AZtCLz3j7QFiOk_w_LiPyXeHpp1gGxxe9xVoIHS28zHco9vaHlsYXzhCn_nLdrHKNsXrejHfZJ4R3mXcqFo4yw0XTFRlqRivHVhqDa0kLRkRklYCnOJCu8ookFJaZ7XoVwLnfIQe_rseAHan5BubzrvLGfwPPPNXfQ
ContentType Conference Proceeding
DBID 6IE
6IH
CBEJK
RIE
RIO
DOI 10.1109/INFOCOM42981.2021.9488906
DatabaseName IEEE Electronic Library (IEL) Conference Proceedings
IEEE Proceedings Order Plan (POP) 1998-present by volume
IEEE Xplore All Conference Proceedings
IEEE Electronic Library (IEL)
IEEE Proceedings Order Plans (POP) 1998-present
DatabaseTitleList
Database_xml – sequence: 1
  dbid: RIE
  name: IEEE Electronic Library (IEL)
  url: https://ieeexplore.ieee.org/
  sourceTypes: Publisher
DeliveryMethod fulltext_linktorsrc
Discipline Engineering
EISBN 9781665403252
166540325X
EISSN 2641-9874
EndPage 10
ExternalDocumentID 9488906
Genre orig-research
GroupedDBID 6IE
6IF
6IH
6IK
6IL
6IM
6IN
AAJGR
AAWTH
ABLEC
ACGFS
ADZIZ
ALMA_UNASSIGNED_HOLDINGS
BEFXN
BFFAM
BGNUA
BKEBE
BPEOZ
CBEJK
CHZPO
IEGSK
IJVOP
IPLJI
M43
OCL
RIE
RIL
RIO
ID FETCH-LOGICAL-i203t-397f4da393424cbb723fdea1a91c61b20461c4ed7348dc97e666ada84848be333
IEDL.DBID RIE
ISICitedReferencesCount 83
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000702210400239&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
IngestDate Wed Aug 27 02:39:50 EDT 2025
IsPeerReviewed false
IsScholarly true
Language English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-i203t-397f4da393424cbb723fdea1a91c61b20461c4ed7348dc97e666ada84848be333
PageCount 10
ParticipantIDs ieee_primary_9488906
PublicationCentury 2000
PublicationDate 2021-May-10
PublicationDateYYYYMMDD 2021-05-10
PublicationDate_xml – month: 05
  year: 2021
  text: 2021-May-10
  day: 10
PublicationDecade 2020
PublicationTitle Annual Joint Conference of the IEEE Computer and Communications Societies
PublicationTitleAbbrev INFOCOM
PublicationYear 2021
Publisher IEEE
Publisher_xml – name: IEEE
SSID ssj0008726
Score 2.5355215
Snippet The conventional federated learning (FedL) architecture distributes machine learning (ML) across worker devices by having them train local models that are...
SourceID ieee
SourceType Publisher
StartPage 1
SubjectTerms Conferences
Device-to-device communication
Machine learning
Network topology
Servers
Training
Wireless networks
Title Device Sampling for Heterogeneous Federated Learning: Theory, Algorithms, and Implementation
URI https://ieeexplore.ieee.org/document/9488906
WOSCitedRecordID wos000702210400239&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LS8NAEF5qEdGLj1Z8s4LHbptkt93Em1RDPdgWVOhBKPuYtAWbSJv6-51NQ1XwIrksCUtgZ7LzTXa-bwi5UR5GeTQ248KzzGF0pnhbMCXQm8EmRmlTNJuQ_X44GkXDCmlsuDAAUBSfQdMNi7N8m5mV-1XWitDbIqevvSVlZ83V2uy6oQw6O-S61NBsPfbjQXfwhLtt6LLAwG-Wk391USmCSLz_v9cfkPo3G48ON3HmkFQgPSJ7P4QEa-TtHtwXT5-VKxBPJxShKO25SpcMHQQwu6exU41AYGlpKak6uaVrYn6D3r1PssUsn86XDapSSwvJ4HnJSkrr5DV-eOn2WNk3gc0Cj-cMIUYirOIRF4EwWsuAJxaUryLfdHwdOI11I8A6YRtrIgmYwiirQoGXBs75MammWQonhHLEI55M8DGAUGGodeArLSPtmzYaX56Smlun8cdaGmNcLtHZ37fPya4zBSvETy9INV-s4JJsm898tlxcFfb8ApRsooY
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1bS8MwFA5DxcuLl028G8HHZWuTbG19k2nZcOsGTtiDMHI5nQPXytb5-026MhV8kb6ElkDJOc35TnO-7yB0KxwT5Y2xCeOOJhajE8EanAhuvBl0rIRUebMJL4r80SgYlFB1zYUBgLz4DGp2mJ_l61Qt7a-yemC8LbD62psNzqmzYmut913fo81tdFOoaNY7Udhv9Xtmv_VtHkjdWjH9Vx-VPIyE-_97gQNU-ebj4cE60hyiEiRHaO-HlGAZvT6A_ebxs7Al4skEGzCK27bWJTUuAia_x6HVjTDQUuNCVHVyh1fU_Cq-f5-k82n2NltUsUg0zkWDZwUvKamgl_Bx2GqTonMCmVKHZcSAjJhrwQLGKVdSepTFGoQrAlc1XUmtyrrioK20jVaBByaJEVr43FwSGGPHaCNJEzhBmBlE4nixeQzAhe9LSV0hvUC6qmHM752isl2n8cdKHGNcLNHZ37ev0U572OuOu53o6RztWrOQXAr1Am1k8yVcoi31mU0X86vctl81oaXN
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Abook&rft.genre=proceeding&rft.title=Annual+Joint+Conference+of+the+IEEE+Computer+and+Communications+Societies&rft.atitle=Device+Sampling+for+Heterogeneous+Federated+Learning%3A+Theory%2C+Algorithms%2C+and+Implementation&rft.au=Wang%2C+Su&rft.au=Lee%2C+Mengyuan&rft.au=Hosseinalipour%2C+Seyyedali&rft.au=Morabito%2C+Roberto&rft.date=2021-05-10&rft.pub=IEEE&rft.eissn=2641-9874&rft.spage=1&rft.epage=10&rft_id=info:doi/10.1109%2FINFOCOM42981.2021.9488906&rft.externalDocID=9488906