Compressed Per-Player Local Methods for Multiplayer Federated Learning: Exact Convergence via Increment Sharing
Abstract
Multiplayer Federated Learning (MpFL), introduced by Yoon et al. (2025), extends classical federated learning to settings where clients are rational players in a game with individual objectives and strategies, seeking an equilibrium. A natural baseline is Per-Player Local SGD (PEARL-SGD), where each player performs several local SGD steps before exchanging strategies through a central server. However, the original PEARL-SGD formulation does not extend cleanly to compressed communication. Directly compressing the players' strategies can prevent convergence to the equilibrium, even with deterministic gradients. To address this issue, we reformulate PEARL-SGD so that players communicate local strategy increments rather than the strategies themselves. Without compression, the two formulations are equivalent, but with compression, increment sharing preserves exact convergence, allowing us to design the communication-efficient compressed PEARL-SGD (C-PEARL-SGD). Beyond compression, we also consider variance reduction as another axis of improvement, introducing PEARL-SVRG and its compressed variant C-PEARL-SVRG. Under the original MpFL assumptions, we establish convergence guarantees for all variants through a unified parametric inequality. Numerical experiments, including distributed spectrum sensing, demonstrate the practical benefits of our methods over vanilla PEARL-SGD.
Related
- Non-Coherent Over-the-Air Federated Learning: Protocol, Convergence, and Device Scheduling
- Superfast Spectral Compressed Sensing with Quadratic Convergence
- Global and Unconditional R-Linear Convergence of Rank-Compressed Weighted LMSD Sweeps for Strictly Convex Quadratics
- Active Learning for Conditional Generative Compressed Sensing
- Compressed Multiscale Kernel Collocation and Physics-Informed Kernel Learning for Elliptic BVPs
BibTeX
@misc{balani2026compressed,
title = {{Compressed Per-Player Local Methods for Multiplayer Federated Learning: Exact Convergence via Increment Sharing}},
author = {Bhavika Balani and TaeHo Yoon and Nicolas Loizou},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.08686},
eprint = {2611.08686},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.08686},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Compressed Per-Player Local Methods for Multiplayer Federated Learning: Exact Convergence via Increment Sharing
AU - Bhavika Balani
AU - TaeHo Yoon
AU - Nicolas Loizou
PY - 2026
DA - 2026/11/05
PB - arXiv
DO - 10.48550/arxiv.2611.08686
UR - https://arxiv.org/abs/2611.08686
AB - Multiplayer Federated Learning (MpFL), introduced by Yoon et al. (2025), extends classical federated learning to settings where clients are rational players in a game with individual objectives and strategies, seeking an equilibrium. A natural baseline is Per-Player Local SGD (PEARL-SGD), where each player performs several local SGD steps before exchanging strategies through a central server. However, the original PEARL-SGD formulation does not extend cleanly to compressed communication. Directly compressing the players' strategies can prevent convergence to the equilibrium, even with deterministic gradients. To address this issue, we reformulate PEARL-SGD so that players communicate local strategy increments rather than the strategies themselves. Without compression, the two formulations are equivalent, but with compression, increment sharing preserves exact convergence, allowing us to design the communication-efficient compressed PEARL-SGD (C-PEARL-SGD). Beyond compression, we also consider variance reduction as another axis of improvement, introducing PEARL-SVRG and its compressed variant C-PEARL-SVRG. Under the original MpFL assumptions, we establish convergence guarantees for all variants through a unified parametric inequality. Numerical experiments, including distributed spectrum sensing, demonstrate the practical benefits of our methods over vanilla PEARL-SGD.
ER -