dc.contributor.author Zhao, Haoyu dc.contributor.author Li, Zhize dc.contributor.author Richtarik, Peter dc.date.accessioned 2021-08-12T13:21:14Z dc.date.available 2021-08-12T13:21:14Z dc.date.issued 2021-08-10 dc.identifier.uri http://hdl.handle.net/10754/670593 dc.description.abstract Federated Averaging (FedAvg, also known as Local-SGD) (McMahan et al., 2017) is a classical federated learning algorithm in which clients run multiple local SGD steps before communicating their update to an orchestrating server. We propose a new federated learning algorithm, FedPAGE, able to further reduce the communication complexity by utilizing the recent optimal PAGE method (Li et al., 2021) instead of plain SGD in FedAvg. We show that FedPAGE uses much fewer communication rounds than previous local methods for both federated convex and nonconvex optimization. Concretely, 1) in the convex setting, the number of communication rounds of FedPAGE is $O(\frac{N^{3/4}}{S\epsilon})$, improving the best-known result $O(\frac{N}{S\epsilon})$ of SCAFFOLD (Karimireddy et al.,2020) by a factor of $N^{1/4}$, where $N$ is the total number of clients (usually is very large in federated learning), $S$ is the sampled subset of clients in each communication round, and $\epsilon$ is the target error; 2) in the nonconvex setting, the number of communication rounds of FedPAGE is $O(\frac{\sqrt{N}+S}{S\epsilon^2})$, improving the best-known result $O(\frac{N^{2/3}}{S^{2/3}\epsilon^2})$ of SCAFFOLD (Karimireddy et al.,2020) by a factor of $N^{1/6}S^{1/3}$, if the sampled clients $S\leq \sqrt{N}$. Note that in both settings, the communication cost for each round is the same for both FedPAGE and SCAFFOLD. As a result, FedPAGE achieves new state-of-the-art results in terms of communication complexity for both federated convex and nonconvex optimization. dc.publisher arXiv dc.relation.url https://arxiv.org/pdf/2108.04755.pdf dc.rights Archived with thanks to arXiv dc.title FedPAGE: A Fast Local Stochastic Gradient Method for Communication-Efficient Federated Learning dc.type Preprint dc.contributor.department Computer Science Program dc.contributor.department Computer, Electrical and Mathematical Science and Engineering (CEMSE) Division dc.contributor.department Visual Computing Center (VCC) dc.eprint.version Pre-print dc.contributor.institution Princeton University, USA dc.identifier.arxivid 2108.04755 kaust.person Li, Zhize kaust.person Richtarik, Peter refterms.dateFOA 2021-08-12T13:22:09Z
﻿

### Files in this item

Name:
Preprintfile1.pdf
Size:
1.148Mb
Format:
PDF
Description:
Pre-print