Show simple item record

dc.contributor.authorHanzely, Filip
dc.contributor.authorHanzely, Slavomir
dc.contributor.authorHorvath, Samuel
dc.contributor.authorRichtarik, Peter
dc.date.accessioned2020-10-13T12:02:06Z
dc.date.available2020-10-13T12:02:06Z
dc.date.issued2020-10-05
dc.identifier.urihttp://hdl.handle.net/10754/665558
dc.description.abstractIn this work, we consider the optimization formulation of personalized federated learning recently introduced by Hanzely and Richt\'arik (2020) which was shown to give an alternative explanation to the workings of local {\tt SGD} methods. Our first contribution is establishing the first lower bounds for this formulation, for both the communication complexity and the local oracle complexity. Our second contribution is the design of several optimal methods matching these lower bounds in almost all regimes. These are the first provably optimal methods for personalized federated learning. Our optimal methods include an accelerated variant of {\tt FedProx}, and an accelerated variance-reduced version of {\tt FedAvg}/Local {\tt SGD}. We demonstrate the practical superiority of our methods through extensive numerical experiments.
dc.publisherarXiv
dc.relation.urlhttps://arxiv.org/pdf/2010.02372
dc.rightsArchived with thanks to arXiv
dc.titleLower Bounds and Optimal Algorithms for Personalized Federated Learning
dc.typePreprint
dc.contributor.departmentApplied Mathematics and Computational Science Program
dc.contributor.departmentComputer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division
dc.contributor.departmentApplied Mathematics & Computational Sci
dc.contributor.departmentStatistics Program
dc.contributor.departmentStatistics
dc.contributor.departmentComputer Science Program
dc.eprint.versionPre-print
dc.identifier.arxivid2010.02372
kaust.personHanzely, Filip
kaust.personHanzely, Slavomir
kaust.personHorvath, Samuel
kaust.personRichtarik, Peter
refterms.dateFOA2020-10-13T12:02:31Z


Files in this item

Thumbnail
Name:
Preprintfile1.pdf
Size:
1.202Mb
Format:
PDF
Description:
Pre-print

This item appears in the following Collection(s)

Show simple item record