PACE: Propagation-Aware Collaborative Correction for One-Shot Personalized Federated Graph Learning

Ruizhe Huang; Chengran Li; Xiaochuan Shi

cs.LGSep 4, 2026

PACE: Propagation-Aware Collaborative Correction for One-Shot Personalized Federated Graph Learning

Authors: Ruizhe Huang, Chengran Li, Xiaochuan Shi

Organizations: School of Cyber Science and Engineering, Wuhan University

Abstract

Client heterogeneity creates both an opportunity and a risk in personalized federated graph learning. Knowledge held by other subgraphs may complement a receiver's Local model, but an incompatible transfer can override reliable predictions. One-shot communication sharpens this tension because an unsuitable server return cannot be corrected later. We introduce PACE, which treats collaborative knowledge as a compact correction to a complete Local predictor rather than as its replacement. Each client uploads a rank-r update carrier and a diagonal sketch of propagated message moments. The server uses them to construct a propagation-aware, receiver-anchored correction, while the receiver retains its full Local model. Convex negative-log-likelihood calibration (CNLL) then selects one coefficient between Local and External logits using validation nodes; model parameters remain fixed and no feedback is sent. At Rank-6, personalized returns occupy 9.6-17.6% of dense tensor bytes across the six evaluated datasets. The correction receives nonzero weight and improves both Accuracy and weighted-F1 over Local on five datasets; on ogbn-arxiv, CNLL assigns zero predictive weight to the correction and preserves Local predictions exactly. Applying the same CNLL rule to matched baselines on three citation datasets does not account for these gains. The central result is therefore that a small transported correction can augment a complete Local model when receiver evidence supports it while leaving the Local prediction unchanged otherwise.

Figures & tables

Figure 1: Two coupled challenges in one-shot personalized FGL. Compact communication must preserve complementary evidence across clients, while the utility of shared knowledge depends on the receiver and can become negative under mismatch.

Figure 2: PACE pipeline. Each client trains locally and uploads a propagation-aware Rank-6 carrier once. The server forms a self-inclusive RegMean consensus and returns a receiver-anchored Rank-6 correction. Each receiver then calibrates a single convex logit coefficient with validation NLL, without parameter updates, test-label selection, feedback, or additional communication.

Methods	Cora		CiteSeer		PubMed
	Accuracy	W-F1	Accuracy	W-F1	Accuracy	W-F1
Local	79.16 $\pm$ 0.25	79.08 $\pm$ 0.27	65.41 $\pm$ 0.72	64.70 $\pm$ 0.73	84.25 $\pm$ 0.05	84.19 $\pm$ 0.05
FedAvg	33.18 $\pm$ 0.94	20.28 $\pm$ 1.74	72.79 $\pm$ 0.32	69.81 $\pm$ 0.46	78.21 $\pm$ 1.07	76.49 $\pm$ 1.49
FedProx	39.61 $\pm$ 2.45	30.88 $\pm$ 3.61	72.82 $\pm$ 0.32	70.12 $\pm$ 0.34	72.32 $\pm$ 2.21	68.28 $\pm$ 2.37
FedNova	29.84 $\pm$ 0.47	14.15 $\pm$ 0.62	38.33 $\pm$ 2.04	33.11 $\pm$ 3.43	38.33 $\pm$ 3.44	29.41 $\pm$ 5.92
FedRCL	23.57 $\pm$ 8.89	15.44 $\pm$ 6.53	26.31 $\pm$ 3.14	18.02 $\pm$ 3.37	35.70 $\pm$ 8.32	19.27 $\pm$ 6.76

Table 1: Results with 10 clients under Louvain partitioning. Values are five-seed mean

\pm

sample standard deviation in percent. Bold and underline denote the best and second-best distinct displayed means, respectively.

Variant	Cora	CiteSeer	PubMed
Local	79.16	65.41	84.25
Transport ( $\alpha=1$ )	63.46	73.82	83.66
Fixed $\alpha=0.5$	80.09	69.96	84.97
Mean + CNLL	79.19	71.77	84.37
Raw-Activation RegMean + CNLL	80.32	73.38	84.79
Direct Return + CNLL	79.91	73.58	84.80

Table 2: Component ablation at

C=10

over five seeds on the three citation benchmarks. Values are mean Accuracy (%). Bold and underline denote the best and second-best distinct displayed means, respectively.

Figure 3: Rank sensitivity and communication at

C=10

. Dashed lines mark Local; the dotted line marks frozen Rank-6; Dense is the uncompressed control.

Dataset	Mean $\alpha$	$\alpha=0$	Upload	Downlink
Cora	0.655	0.0%	12.1%	10.3%
CiteSeer	0.979	0.0%	11.4%	9.7%
PubMed	0.278	0.0%	13.6%	11.4%
CS	0.512	0.0%	11.2%	9.6%
Computers	0.134	22.0%	13.1%	11.1%
ogbn-arxiv	0.000	100.0%	21.6%	17.6%

Table 3: PACE communication and CNLL utilization at

C=10

. Ratios use complete serialized Rank-6 PACE upload/download bytes relative to dense tensor bytes.

Accuracy
Dataset	Matched method	Score	PACE	Gap
Cora	FedAux-1R+CNLL	79.69	80.34	+0.65
CiteSeer	FedAvg+CNLL	73.11	73.82	+0.71
PubMed	FedTAD+CNLL	84.44	84.79	+0.35
Weighted-F1
Dataset	Matched method	Score	PACE	Gap

Table 4: Matched-CNLL fairness summary at

C=10

over five seeds. For each metric and dataset, the matched method is the highest-scoring baseline after receiving the same receiver-local CNLL calibration; Gap is PACE-CNLL minus its score in percentage points.

Federated Graph Learning

Federated graph learning trains graph models across data holders without centralizing their private nodes, edges, or features.

See all 46 papers

Jun 11, 2026

FedSPC: Shared Parameter Correction for Personalized Federated Learning

Personalized federated learning (PFL) is one of the important approaches in federated learning for addressing statistical heterogeneity while enabling client-specific adaptation. Many PFL methods split the model into shared and personalized parameters, which are jointly trained on each client. However, this creates an optimization issue: shared parameters are updated by clients optimizing different local objectives, which can lead to inconsistent shared updates and weaken the shared representation. To address this problem, we propose Federated Shared Parameter Correction (FedSPC), a modular correction method for PFL. FedSPC applies control-variate correction only to the shared parameters of a given PFL method, while leaving personalized parameters unchanged. It can be integrated into three common PFL settings: shared feature extractors, shared classifiers, and fully shared models with local regularization. Experiments on CIFAR-100 and Tiny-ImageNet with ViT, ResNet-34, and VGG-11 show that FedSPC improves performance across representative PFL methods, including FedPer, FedRep, FedBABU, LG-FedAvg, and Ditto.

Kannanthodath Induchoodan Ajay Menon, Christian Prehofer, Yunfei Xu +1

DENSO AUTOMOTIVE Deutschland GmbH, Germany · Technical University of Munich, Germany · DENSO International America, Inc., United States

Non-IID Federated Learning Federated Learning Heterogeneous Federated Learning Parameter Sharing

Ask PDF

Sep 29, 2026

FedLAFP: Low-Rank Aggregation Meets Full-Rank Personalization in Federated Fine-Tuning

Federated parameter-efficient fine-tuning enables clients to adapt pre-trained models without sharing raw data or communicating the full model, but statistical heterogeneity makes a single global adapter insufficient for personalized prediction. Existing personalized methods typically use the same low-rank structure for both shared and private adaptation, overlooking their distinct requirements for aggregation and personalization. We propose FedLAFP, a role-aware framework that couples a compact, globally aggregated LoRA branch with a client-private, full-rank-capable RandLoRA branch. The shared branch provides an efficient interface for transferring common knowledge, whereas the private branch combines fixed random low-rank bases with learned scaling coefficients to provide expressive client-specific adaptation without additional communication. Client- and layer-specific mixing coefficients jointly fuse the two branches, and only the shared LoRA parameters are exchanged. A controlled linear study supports this role assignment: LoRA yields more aligned client updates and lower aggregation error, while RandLoRA more accurately recovers client-specific residuals. Experiments across four visual recognition benchmarks show that FedLAFP consistently outperforms local-only and federated LoRA baselines, achieving an average personalized accuracy of

86.93\%

and exceeding the best baseline average by

1.30

percentage points.

Mengjun Yi, Huaian Gu, Yinghao Ai +2

State Key Laboratory for Novel Software Technology · School of Artificial Intelligence · School of Computer Science +1

Low-Rank Adaptation Federated Fine-Tuning Federated Learning Parameter-Efficient Fine-Tuning

Ask PDF

May 26, 2026

Separate Aggregation of Split Network for Personalized Federated Learning

Federated learning enables collaborative model training without sharing raw data, but its performance can degrade substantially under heterogeneous client data distributions. A single global model often cannot satisfy diverse client requirements, so personalized federated learning has therefore been explored to improve client specific performance while preserving global generalization. Existing PFL methods often face a fundamental tradeoff in which stronger global sharing can undermine local specialization, whereas stronger local adaptation can lead to overfitting under limited data, label imbalance, and missing class scenarios. In this work, we propose PGFedSplit, a personalized federated learning framework that improves both personalization and global generalization under severe client heterogeneity. PGFedSplit adopts a split architecture and performs adaptive aggregation scheduling tailored to the roles of different model components, enabling stable knowledge sharing while maintaining client specific adaptation. Each client further leverages a mixture of locally extracted representations and synthetic representations generated from server side Gaussian statistics, improving robustness under label imbalance and missing class conditions. Extensive experiments on Fashion MNIST, CIFAR 10, CIFAR 100, and Tiny ImageNet demonstrate consistent improvements over state of the art PFL methods, with stable convergence and superior personalization in highly heterogeneous settings.

Yunseok Kang, Jaeyoung Song

Department of Electronics Engineering, Pusan National University, Republic of Korea

Federated Learning Aggregation Split Federated Learning Federated Learning Heterogeneous Federated Learning

Ask PDF

Abstract

Federated Graph Learning

Explore similar work

FedSPC: Shared Parameter Correction for Personalized Federated Learning

FedLAFP: Low-Rank Aggregation Meets Full-Rank Personalization in Federated Fine-Tuning

Separate Aggregation of Split Network for Personalized Federated Learning