One-shot Distributed Learning via Weighted Averaging with Server-side Information
Abstract
Distributed learning has drawn considerable attention due to its capability of learning from distributed data without aggregating raw observations. However, frequent exchange of model updates can become the main bottleneck of communication. Therefore, one-shot approaches where each worker fits a local model and sends it to the server to aggregate emerge. It is known that averaging can achieve the so-called pooled-data mean squared error (MSE) order of a hypothetical node that has access to data from the server and all the workers, and performs empirical risk minimization (ERM) on it. However, when the number of workers increases too fast compared to the local sample size, nonzero local estimation bias prevents averaging from achieving this order. In this paper, we study whether independent server data can overcome this limitation. We consider m workers with n observations each and a server with p observations, all i.i.d., giving N=mn+p observations in total. Under certain regularity conditions, a pooled-data ERM achieves MSE of order 𝒪(N−1). We analyze deterministic nonnegative averaging of convex-loss M-estimators, with equal worker weights and the server's own estimator included. Under our assumptions, including nonzero local bias of order n−α, α≥1, sample-size weighted averaging achieves the pooled-data MSE order whenever m=𝒪(n2α−1). Above this worker scale, it does so if and only if p=Ω(nm· m/n2α−1). Hence, server-only estimation, whose MSE is of order 𝒪(p−1), can achieve pooled-data order while sample-size weighting fails. Analytical examples and synthetic experiments illustrate when averaging matches the pooled-data MSE order, complemented by CIFAR-10 experiments comparing aggregation strategies with linear and nonlinear classifiers.
Related
- Equilibria and consensus in discrete-time opinion dynamics with mixed weighted-averaging and weighted-median mechanisms
- CoeF-SFL: Preserving Collaborative Server-Client Learning with Enhanced Communication Efficiency
- Universal Distributed Feature Extraction Complementary to Side Information via H-Score
- Distributed Stochastic Approximation Algorithms and Heavy-Tailed Age of Information
- An information-geometric formulation of distributed computation
BibTeX
@misc{tepavcevic2026oneshot,
title = {{One-shot Distributed Learning via Weighted Averaging with Server-side Information}},
author = {Pavle Tepavčević and Dušan Jakovetić and Miloš Savić},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.04700},
eprint = {2611.04700},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.04700},
}
RIS
TY - UNPB
TI - One-shot Distributed Learning via Weighted Averaging with Server-side Information
AU - Pavle Tepavčević
AU - Dušan Jakovetić
AU - Miloš Savić
PY - 2026
DA - 2026/11/03
PB - arXiv
DO - 10.48550/arxiv.2611.04700
UR - https://arxiv.org/abs/2611.04700
AB - Distributed learning has drawn considerable attention due to its capability of learning from distributed data without aggregating raw observations. However, frequent exchange of model updates can become the main bottleneck of communication. Therefore, one-shot approaches where each worker fits a local model and sends it to the server to aggregate emerge. It is known that averaging can achieve the so-called pooled-data mean squared error (MSE) order of a hypothetical node that has access to data from the server and all the workers, and performs empirical risk minimization (ERM) on it. However, when the number of workers increases too fast compared to the local sample size, nonzero local estimation bias prevents averaging from achieving this order. In this paper, we study whether independent server data can overcome this limitation. We consider m workers with n observations each and a server with p observations, all i.i.d., giving N=mn+p observations in total. Under certain regularity conditions, a pooled-data ERM achieves MSE of order 𝒪(N^−1). We analyze deterministic nonnegative averaging of convex-loss M-estimators, with equal worker weights and the server's own estimator included. Under our assumptions, including nonzero local bias of order n^−α, α≥1, sample-size weighted averaging achieves the pooled-data MSE order whenever m=𝒪(n^2α−1). Above this worker scale, it does so if and only if p=Ω(nm· m/n^2α−1). Hence, server-only estimation, whose MSE is of order 𝒪(p^−1), can achieve pooled-data order while sample-size weighting fails. Analytical examples and synthetic experiments illustrate when averaging matches the pooled-data MSE order, complemented by CIFAR-10 experiments comparing aggregation strategies with linear and nonlinear classifiers.
ER -