FedHQ: Hybrid Runtime Quantization for Federated Learning

Zihao Zheng; Ziyao Wang; Jiawei Yang; Hangyu Cao; Zhihao Mao; Maoliang Li; Jiayu Chen; Hailong Zou

cs.LGMay 17, 2025

FedHQ: Hybrid Runtime Quantization for Federated Learning

Authors: Zihao Zheng, Ziyao Wang, Jiawei Yang, Hangyu Cao, Zhihao Mao, Maoliang Li, Jiayu Chen, Hailong Zou,

Organizations: Peking University · University of Maryland

Abstract

Federated Learning (FL) must balance training efficiency and model accuracy across heterogeneous clients. Post-training quantization (PTQ) avoids the extra computation of quantization-aware training (QAT), but can incur greater accuracy loss. We propose FedHQ, a hybrid quantization framework that assigns PTQ or QAT according to clients' hardware and data characteristics. Hardware-based speed estimates and local data-distribution indicators guide coarse-grained global initialization, followed by fine-grained training-based adjustment of selected client strategies. Experiments on three image-classification datasets show that FedHQ achieves up to 2.47x training speedup over FedAQT and up to 11.15 percentage points higher accuracy than FedFQ.

Figures & tables

Fig. 1: Quantization workflows in FL: (a) QAT; (b) PTQ.

Fig. 2: Motivating quantization study: (a) uniform and hybrid allocations; (b) resource constraints and batch-size sensitivity; (c) convergence on CIFAR-10. Device icons illustrate deployment constraints.

Fig. 3: Geometric view of client-level strategy allocation. The horizontal axis represents a data-derived indicator, not measured accuracy; the small-area region is an allocation heuristic.

Fig. 4: Overview of FedHQ initialization and training-based refinement.

Fig. 5: Accuracy Trajectories on CIFAR-10, CIFAR-100, and SVHN.

Task	FedAQT		FedFQ		FedRH		FedHQ
Task	Acc.	Speed	Acc.	Speed	Acc.	Speed	Acc.	Speed
CIFAR-10	67.33	1.00 $\times$	64.46	5.43 $\times$	66.65	1.64 $\times$	67.78	1.80 $\times$
CIFAR-100	42.48	1.00 $\times$	41.32	5.28 $\times$	42.36	2.09 $\times$	46.35	2.47 $\times$
SVHN	82.85	1.00 $\times$	81.86	2.85 $\times$	82.21	1.45 $\times$	93.01	1.50 $\times$

TABLE I: Final accuracy (%) and training speedup relative to FedAQT.

Fig. 6: FedHQ ablations: (a) allocation strategies; (b) indicator validity; (c) robustness under heterogeneous settings; (d) refinement efficiency.

Dataset	$\xi=0.3$	$\xi=0.2$ (Selected)	$\xi=0.1$
CIFAR-10	210.4 / 64.31	221.0 / 67.78	240.5 / 66.42
CIFAR-100	153.8 / 46.87	150.7 / 46.35	179.3 / 47.15
SVHN	151.2 / 92.46	163.1 / 93.01	173.8 / 82.21

TABLE II: Effect of

\xi

on training time (min) / accuracy (%).

Federated Learning

Federated learning trains shared models across data holders without requiring them to pool their raw data.

See all 943 papers

Quantization in Federated Learning: Methods, Challenges and Future Directions
Jun 25, 2026Farwa Ikram, Dipanwita Thakur, Antonella Guzzo +1Neural Network Quantization Federated Learning
DIMES - University of Calabria, Italy
Quantization Impact on the Accuracy and Communication Efficiency Trade-off in Federated Learning for Aerospace Predictive Maintenance
Apr 9, 2026Abdelkarim LoukiliGradient Quantization Federated Learning
Enhanced Privacy and Communication Efficiency in Non-IID Federated Learning with Adaptive Quantization and Differential Privacy
Apr 25, 2026Emre Ardıç, Yakup GençNon-IID Federated Learning Differentially Private Federated Learning
Department of Computer Engineering, Gebze Technical University, Gebze, Kocaeli 41400, Türkiye

Jun 25, 2026J/K move · Enter open · S save

Quantization in Federated Learning: Methods, Challenges and Future Directions

Farwa Ikram, Dipanwita Thakur, Antonella Guzzo, Giancarlo Fortino

DIMES - University of Calabria, Italy

Neural Network Quantization Federated Learning

Open paper AskPDF arXiv

Federated Learning (FL) has become a foundational paradigm for privacy-preserving distributed intelligence, yet its scalability remains fundamentally constrained by communication bottlenecks, device heterogeneity, and the challenges of training under statistically non-IID data. Quantization is one of the most effective mechanisms for mitigating these limitations, reducing both uplink/downlink payloads and on-device computation. This paper provides the first FL-centric systematic review of quantization, introducing a novel taxonomy organized around FL-specific dimensions, including client heterogeneity, aggregation consistency, communication-scheduling adaptation, non-IID robustness, privacy/security integration, and hardware/energy co-optimization. Beyond cataloging existing methods, we analyze how quantization interacts with core FL behaviors such as client drift, partial participation, convergence stability, secure aggregation, and differential privacy. We further identify cross-method insights, open research gaps, and design guidelines for practitioners deploying quantized FL on mobile, IoT, and edge platforms. This survey thus establishes quantization not merely as a compression technique, but as a fundamental systems component shaping the performance, robustness, and practicality of modern FL.

Abstract

Federated Learning

Explore similar work

Quantization in Federated Learning: Methods, Challenges and Future Directions

Quantization Impact on the Accuracy and Communication Efficiency Trade-off in Federated Learning for Aerospace Predictive Maintenance

Enhanced Privacy and Communication Efficiency in Non-IID Federated Learning with Adaptive Quantization and Differential Privacy