Fedvarp: Tackling the variance due to partial client participation in federated learning

Divyansh Jhunjhunwala, Pranay Sharma, Aushim Nagarkatti, Gauri Joshi
Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence, PMLR 180:906-916, 2022.

Abstract

Data-heterogeneous federated learning (FL) systems suffer from two significant sources of convergence error: 1) client drift error caused by performing multiple local optimization steps at clients, and 2) partial client participation error caused by the fact that only a small subset of the edge clients participate in every training round. We find that among these, only the former has received significant attention in the literature. To remedy this, we propose FedVARP, a novel variance reduction algorithm applied at the server that eliminates error due to partial client participation. To do so, the server simply maintains in memory the most recent update for each client and uses these as surrogate updates for the non-participating clients in every round. Further, to alleviate the memory requirement at the server, we propose a novel clustering-based variance reduction algorithm ClusterFedVARP. Unlike previously proposed methods, both FedVARP and ClusterFedVARP do not require additional computation at clients or communication of additional optimization parameters. Through extensive experiments, we show that FedVARP outperforms state-of-the-art methods, and ClusterFedVARP achieves performance comparable to FedVARP with much less memory requirements.

Cite this Paper


BibTeX
@InProceedings{pmlr-v180-jhunjhunwala22a, title = {Fedvarp: Tackling the variance due to partial client participation in federated learning}, author = {Jhunjhunwala, Divyansh and Sharma, Pranay and Nagarkatti, Aushim and Joshi, Gauri}, booktitle = {Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence}, pages = {906--916}, year = {2022}, editor = {Cussens, James and Zhang, Kun}, volume = {180}, series = {Proceedings of Machine Learning Research}, month = {01--05 Aug}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v180/jhunjhunwala22a/jhunjhunwala22a.pdf}, url = {https://proceedings.mlr.press/v180/jhunjhunwala22a.html}, abstract = {Data-heterogeneous federated learning (FL) systems suffer from two significant sources of convergence error: 1) client drift error caused by performing multiple local optimization steps at clients, and 2) partial client participation error caused by the fact that only a small subset of the edge clients participate in every training round. We find that among these, only the former has received significant attention in the literature. To remedy this, we propose FedVARP, a novel variance reduction algorithm applied at the server that eliminates error due to partial client participation. To do so, the server simply maintains in memory the most recent update for each client and uses these as surrogate updates for the non-participating clients in every round. Further, to alleviate the memory requirement at the server, we propose a novel clustering-based variance reduction algorithm ClusterFedVARP. Unlike previously proposed methods, both FedVARP and ClusterFedVARP do not require additional computation at clients or communication of additional optimization parameters. Through extensive experiments, we show that FedVARP outperforms state-of-the-art methods, and ClusterFedVARP achieves performance comparable to FedVARP with much less memory requirements.} }
Endnote
%0 Conference Paper %T Fedvarp: Tackling the variance due to partial client participation in federated learning %A Divyansh Jhunjhunwala %A Pranay Sharma %A Aushim Nagarkatti %A Gauri Joshi %B Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2022 %E James Cussens %E Kun Zhang %F pmlr-v180-jhunjhunwala22a %I PMLR %P 906--916 %U https://proceedings.mlr.press/v180/jhunjhunwala22a.html %V 180 %X Data-heterogeneous federated learning (FL) systems suffer from two significant sources of convergence error: 1) client drift error caused by performing multiple local optimization steps at clients, and 2) partial client participation error caused by the fact that only a small subset of the edge clients participate in every training round. We find that among these, only the former has received significant attention in the literature. To remedy this, we propose FedVARP, a novel variance reduction algorithm applied at the server that eliminates error due to partial client participation. To do so, the server simply maintains in memory the most recent update for each client and uses these as surrogate updates for the non-participating clients in every round. Further, to alleviate the memory requirement at the server, we propose a novel clustering-based variance reduction algorithm ClusterFedVARP. Unlike previously proposed methods, both FedVARP and ClusterFedVARP do not require additional computation at clients or communication of additional optimization parameters. Through extensive experiments, we show that FedVARP outperforms state-of-the-art methods, and ClusterFedVARP achieves performance comparable to FedVARP with much less memory requirements.
APA
Jhunjhunwala, D., Sharma, P., Nagarkatti, A. & Joshi, G.. (2022). Fedvarp: Tackling the variance due to partial client participation in federated learning. Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research 180:906-916 Available from https://proceedings.mlr.press/v180/jhunjhunwala22a.html.

Related Material