Federated Learning of Binary Neural Networks: Enabling Low-Cost Inference

Nitin Priyadarshini Shankar; Soham Lahiri; Sheetal Kalyani; Saurav Prakash

cs.LGMar 16, 2026

Federated Learning of Binary Neural Networks: Enabling Low-Cost Inference

Authors: Nitin Priyadarshini Shankar, Soham Lahiri, Sheetal Kalyani, Saurav Prakash

Organizations: Department of Electrical Engineering, Indian Institute of Technology Madras · Department of Computer Science and Engineering, Jadavpur University

Abstract

Federated Learning (FL) preserves privacy by distributing training across devices. However, using DNNs is computationally intensive at the low-powered edge during inference. Edge deployment demands models that simultaneously optimize memory footprint and computational efficiency, a dilemma where conventional DNNs fail by exceeding resource limits. Traditional post-training binarization reduces model size but suffers from severe accuracy loss due to quantization errors. To address these challenges, we propose FedBNN, a rotation-aware binary neural network framework that learns binary representations directly during local training. By encoding each weight as a single bit $\{+1, -1\}$ instead of a $32$ -bit float, FedBNN shrinks the model footprint, significantly reducing runtime (during inference) FLOPs and memory requirements in comparison to federated methods using real models. Evaluations across multiple benchmark datasets demonstrate that FedBNN significantly reduces resource consumption while performing similarly to existing federated methods using real-valued models.

Figures & tables

Method	Dataset	Accuracy			FLOPs	Memory	Binarized Accuracy
	(Model)	IID	Non -IID 1	Non -IID 2		(MB)	IID	Non -IID 1	Non -IID 2
FedAvg		$\mathbf{92.24}$	$\mathbf{91.44}$	$\mathbf{89.28}$	$2.02\times 10^{7}$	$1.5635$	$53.42$	$63.68$	$54.72$
FedBAT	FMNIST	$89.12$	$87.66$	$85.56$	$2.02\times 10^{7}$	$1.5635$	$14.34$	$16.98$	$8.0$
FedMUD	(CNN4)	$89.60$	$88.60$	$86.00$	$2.02\times 10^{7}$	$1.6291$	$63.2$	$66.5$	$66.08$
FedBNN		$88.46$	$87.76$	$83.38$	$\mathbf{3.48\times 10^{5}}$	$\mathbf{0.0489}$	$\mathbf{88.46}$	$\mathbf{87.76}$	$\mathbf{83.38}$
FedAvg		$\mathbf{92.10}$	$\mathbf{90.60}$	$\mathbf{89.34}$	$3.00\times 10^{7}$	$1.5965$	$28.01$	$22.56$	$16.92$

Table 1: Performance comparison for

N_{c}=100

. The FLOPs and memory values are calculated during runtime. Binarized accuracy refers to the model’s performance after the weights and activations have been binarized.

Method	Dataset (Model)	IID	Non-IID 1	Non-IID 2
FedBNN	FMNIST	$88.46$	$\mathbf{87.76}$	$\mathbf{83.38}$
FedBNN ( $\beta=1$ , $\lambda=1$ )	(CNN4)	$\mathbf{88.54}$	$87.68$	$82.48$
FedBNN	SVHN	$\mathbf{86.89}$	$\mathbf{85.63}$	$\mathbf{84.40}$
FedBNN ( $\beta=1$ , $\lambda=1$ )	(CNN4)	$86.77$	$85.38$	$84.13$
FedBNN	CIFAR10	$\mathbf{89.95}$	$\mathbf{82.84}$	$\mathbf{73.82}$
FedBNN ( $\beta=1$ , $\lambda=1$ )	(ResNet10)	$89.34$	$82.82$	$68.54$

Table 2: Ablation study for

N_{c}=100

comparing with and without server aggregation.

Appendix figures & tables16 assets

Supplementary material from the paper’s appendix.

Appendix

Method	Dataset	Accuracy
	(Model)	IID	Non-IID 1	Non-IID 2
FedBNN	FMNIST	$\mathbf{88.46}$	$\mathbf{87.76}$	$\mathbf{83.38}$
Aggregating $\tilde{\mathbf{w}}$	(CNN4)	$87.88$	$87.24$	$80.40$
FedBNN	SVHN	$\mathbf{86.89}$	$\mathbf{85.63}$	$\mathbf{84.40}$
Aggregating $\tilde{\mathbf{w}}$	(CNN4)	$86.25$	$84.74$	$80.40$
FedBNN	CIFAR10	$\mathbf{89.95}$	$\mathbf{82.84}$	$\mathbf{73.82}$

Appendix

Table 3: Ablation study under different aggregation and binarization strategies for

N_{c}=100

Method	Dataset	Accuracy
	(Model)	IID	Non-IID 1	Non-IID 2
FedBNN	CIFAR10	$\mathbf{72.08}$	$\mathbf{67.08}$	$\mathbf{63.00}$
FedBNN ( $\beta=1$ , $\lambda=1$ )	(ConvNext-Tiny)	$70.78$	$61.78$	$58.50$

Appendix

Table 4: Ablation study for

N_{c}=100

comparing with and without server aggregation.

Method	Dataset	(Model)	Accuracy			FLOPs	Memory
			IID	Non-IID 1	Non-IID 2		(MB)
FedBNN		ResNet10	${89.95}$	${82.84}$	$\mathbf{73.82}$	$1.11\times 10^{7}$	$0.61$
FedAvg		ResNet10	$\mathbf{90.86}$	$\mathbf{86.28}$	${70.62}$	$4.40\times 10^{8}$	$19.62$
FedAvg	CIFAR10	ResNet10 (FLOPs matched)	${84.74}$	${81.06}$	${66.16}$	$1.12\times 10^{7}$	${0.49}$
FedAvg		ResNet10 (Memory matched)	${86.60}$	${81.50}$	${68.78}$	$1.35\times 10^{7}$	${0.59}$
FedBNN		ResNet18	${46.54}$	${43.00}$	${45.74}$	$9.05\times 10^{7}$	$1.41$

Appendix

Table 5: Performance comparison for

N_{c}=100

. The FLOPs and memory values are calculated during runtime.

Method	Dataset	Model	No of parameters for $\mathbf{w}$	No of parameters for $\mathbf{R}_{1},\mathbf{R}_{2}$	Overhead Percentage
FedBNN	FMNIST	CNN4	$387,360$	$4,070$	$1.05\%$
	SVHN	CNN4	$387,936$	$4,142$	$1.07\%$
	CIFAR10	ResNet10	$4,890,624$	$27,552$	$0.56\%$
	FEMNIST	ResNet18	$11,157,504$	$62,272$	$0.56\%$
	TinyImageNet	ResNet18	$11,157,504$	$62,272$	$0.56\%$
	CIFAR10	ConvNext-Tiny	$27,751,392$	$78,084$	$0.28\%$

Appendix

Table 6: Parameter overhead of the rotation matrices

\mathbf{R}_{1}

and

\mathbf{R}_{2}

in FedBNN. Across all datasets and architectures, the additional parameters remain below

1\%

of the model weights, demonstrating negligible overhead (in comparison to FedAvg) and good scalability to larger models.

Panel of Figure 2 — Figure 2: Performance comparison between FedAvg and FedBNN on FMNIST using the CNN4 architecture under IID, Non-IID1 (Dirichlet $\alpha=0.3$ ), and Non-IID2 (label-skew) data partitions.

Panel of Figure 3 — Figure 3: Performance comparison between FedAvg and FedBNN on SVHN using the CNN4 architecture under IID, Non-IID1 (Dirichlet $\alpha=0.3$ ), and Non-IID2 (label-skew) data partitions.

Panel of Figure 4 — Figure 4: Performance comparison between FedAvg and FedBNN on CIFAR10 using the ResNet10 architecture under IID, Non-IID1 (Dirichlet $\alpha=0.3$ ), and Non-IID2 (label-skew) data partitions.

Panel of Figure 5 — Figure 5: Performance comparison between FedAvg and FedBNN on CIFAR10 using the ConvNeXt-Tiny architecture under IID, Non-IID1 (Dirichlet $\alpha=0.3$ ), and Non-IID2 (label-skew) data partitions.

Panel of Figure 6 — Figure 6: Performance comparison between FedAvg and FedBNN on TinyImageNet using the ResNet18 architecture under IID, Non-IID1 (Dirichlet $\alpha=0.3$ ), and Non-IID2 (label-skew) data partitions.

Panel of Figure 7 — Figure 7: Performance comparison between FedAvg and FedBNN on FEMNIST using the ResNet18 architecture under IID, Non-IID1 (Dirichlet $\alpha=0.3$ ), and Non-IID2 (label-skew) data partitions.

Panel of Figure 8 — Figure 8: Hyperparameter sensitivity on FMNIST under IID, Non-IID1, and Non-IID2.

Panel of Figure 9 — Figure 9: Hyperparameter sensitivity on SVHN under IID, Non-IID1, and Non-IID2.

Panel of Figure 10 — Figure 10: Hyperparameter sensitivity on CIFAR10 (ResNet10) under IID, Non-IID1, and Non-IID2 data heterogeneity settings.

Panel of Figure 11 — Figure 11: Sensitivity analysis of the hyperparameters $(1-\alpha)$ , $\alpha\beta$ , $\alpha(1-\beta)$ , and $\lambda$ on CIFAR10 (ConvNeXT-Tiny) under three data heterogeneity settings: IID, Non-IID1 (LabelCnt=0.3), and Non-IID2 (Dirichlet=0.3).

Panel of Figure 12 — Figure 12: Hyperparameter sensitivity on TinyImageNet under IID, Non-IID1, and Non-IID2 settings.

Panel of Figure 13 — Figure 13: Hyperparameter sensitivity on FEMNIST under IID, Non-IID1, and Non-IID2 settings.

Explore similar work

QuantFL: Sustainable Federated Learning for Edge IoT via Pre-Trained Model Quantisation
Mar 18, 2026Charuka Herath, Yogachandran Rahulamathavan, Varuna De Silva +1Federated Learning Neural Network Quantization
Institute of Digital Technologies, Loughborough University, UK
Collate: Collaborative Neural Network Learning for Latency-Critical Edge Systems
Jul 9, 2026Shuo Huai, Di Liu, Hao Kong +5Efficient Neural Network Inference Federated Learning Aggregation
School of Computer Science and Engineering, Nanyang Technological University, Singapore · HP-NTU Digital Manufacturing Corporate Lab, Nanyang Technological University, Singapore · HP Inc., Palo Alto, California, USA
Federated Lightweight Fine-Tuning
Jul 20, 2026Radhakrishna Achanta, Will ReedFederated Learning Aggregation Federated Fine-Tuning
Cisco Systems Inc., USA

Mar 18, 2026J/K move · Enter open · S save

QuantFL: Sustainable Federated Learning for Edge IoT via Pre-Trained Model Quantisation

Charuka Herath, Yogachandran Rahulamathavan, Varuna De Silva, Sangarapillai Lambotharan

Institute of Digital Technologies, Loughborough University, UK

Federated Learning Neural Network Quantization Internet of Things

Open paper AskPDF arXiv

Federated Learning (FL) enables privacy-preserving intelligence on Internet of Things (IoT) devices but incurs a significant carbon footprint due to the high energy cost of frequent uplink transmission. While pre-trained models are increasingly available on edge devices, their potential to reduce the energy overhead of fine-tuning remains underexplored. In this work, we propose QuantFL, a sustainable FL framework that leverages pre-trained initialisation to enable aggressive, computationally lightweight quantisation. We demonstrate that pre-training naturally concentrates update statistics, allowing us to use memory-efficient bucket quantisation without the energy-intensive overhead of complex error-feedback mechanisms. On MNIST and CIFAR-100, QuantFL reduces total communication by 40% (

\simeq40\%

total-bit reduction with full-precision downlink;

\geq80\%

on uplink or when downlink is quantised) while matching or exceeding uncompressed baselines under strict bandwidth budgets; BU attains 89.00% (MNIST) and 66.89% (CIFAR-100) test accuracy with orders of magnitude fewer bits. We also account for uplink and downlink costs and provide ablations on quantisation levels and initialisation. QuantFL delivers a practical, "green" recipe for scalable training on battery-constrained IoT networks.

Abstract

Explore similar work

QuantFL: Sustainable Federated Learning for Edge IoT via Pre-Trained Model Quantisation

Collate: Collaborative Neural Network Learning for Latency-Critical Edge Systems

Federated Lightweight Fine-Tuning