arXiv cs.AI· Shudi Weng, Chao Ren, Ming Xiao, Mikael Skoglund·· 6 天前AI 评分22
FedACS：面向异构联邦学习的最优高效客户端采样方法
Heterogeneity-Aware Client Sampling for Optimal and Efficient Federated Learning
AI 导读
针对联邦学习中通信与计算异构性共同导致的目标不一致问题，研究者首次给出通用异构联邦学习的统一理论分析，揭示两类异构性扭曲优化轨迹的不同机制。据此提出客户端采样方法 FedACS，可消除各类目标不一致，理论上以 O(1/√R) 速率收敛到正确最优解。实验显示其优于 SOTA 及类别专用基线 4.3%-36%，通信成本降低 22%-89%，计算负载降低 14%-105%。
正文 · 原文
Abstract:Federated learning (FL) commonly involves clients with diverse communication and computational capabilities. Such heterogeneity can significantly distort the optimization dynamics and lead to objective inconsistency, where the global model converges to an incorrect stationary point potentially far from the pursued optimum. Despite its critical impact, the joint effect of communication and computation heterogeneity has remained largely unexplored, due to the intrinsic complexity of their interaction. In this paper, we reveal the fundamentally distinct mechanisms through which heterogeneous communication and computation drive inconsistency in FL. To the best of our knowledge, this is the first unified theoretical analysis of general heterogeneous FL, offering a principled understanding of how these two forms of heterogeneity jointly distort the optimization trajectory under arbitrary choices of local solvers. Motivated by these insights, we propose Federated Heterogeneity-Aware Client Sampling, FedACS, a universal method to eliminate all types of objective inconsistency. We theoretically prove that FedACS converges to the correct optimum at a rate of $O(1/\sqrt{R})$, even in dynamic heterogeneous environments. Extensive experiments across multiple datasets show that FedACS outperforms state-of-the-art and category-specific baselines by 4.3%-36%, while reducing communication costs by 22%-89% and computation loads by 14%-105%, respectively.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2505.11304 [cs.LG] |
| (or arXiv:2505.11304v4 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2505.11304 arXiv-issued DOI via DataCite |
Submission history
From: Shudi Weng [view email]
[v1]
Fri, 16 May 2025 14:31:36 UTC (1,124 KB)
[v2]
Mon, 16 Feb 2026 06:46:40 UTC (585 KB)
[v3]
Sun, 22 Feb 2026 05:58:59 UTC (580 KB)
[v4]
Fri, 9 Oct 2026 01:27:04 UTC (1,377 KB)
来源：arXiv cs.AI · arxiv.org