PaSta: Noisy Node Classification with Partial Label Learning

Yujing Liu; Yixin Liu; Yu Zheng; Yue Tan; Alan Wee-Chung Liew; Shirui Pan

cs.LGAug 26, 2026

PaSta: Noisy Node Classification with Partial Label Learning

Authors: Yujing Liu, Yixin Liu, Yu Zheng, Yue Tan, Alan Wee-Chung Liew, Shirui Pan

Organizations: School of Information and Communication Technology, Griffith University, Brisbane, Australia

Abstract

Noisy node classification problem is a fundamental yet challenging task for real-world graph-related web services, where node labels are often corrupted or unreliable due to weak supervision or automatic annotation. However, existing methods typically train models based on one-hot labels, which not only makes models susceptible to overfitting on noisy labels, but also leads to error accumulation after pseudo-label-guided enhancement. In this paper, we propose a novel Partial label-based Self-training framework (PaSta for short) that leverages partial label learning technique to overcome the limitations of existing methods. Specifically, PaSta first trains multiple annotators to comprehensively capture the class distribution of nodes and aggregates their predictions to construct high-quality partial labels. Subsequently, we design a partial label-based classification model with two well-crafted loss functions to guide the model learning at both label and representation spaces. To further enhance the robustness against noisy labels, we introduce a self-training strategy where the labels refined by partial label learning are then used to further optimize the annotators in a closed-loop iterative manner. Extensive experiments on five datasets demonstrate that, compared with existing state-of-the-art methods, PaSta achieves an average improvement of 1.1% in classification performance under various noise settings.

Figures & tables

Fig. 1 : Sketch maps of noisy label and partial label learning. While noisy labels tend to drive the model toward confident but potentially incorrect predictions, partial labels help the model produce ambiguous yet more accurate predictions.

Fig. 2 : The framework of the proposed PaSta . On the left, we leverage diverse self-supervised graph learning algorithms to construct annotators and generate consensus labels. On the right, we introduce $\mathcal{L}_{pasim}$ and $\mathcal{L}_{vace}$ to guide the classifier in learning robust patterns from the consensus labels. The trained classifier is then used to filter and expand dataset labels, which are subsequently used to further optimize the annotators’ predictors. Finally, the two modules are iteratively optimized.

Methods	Uniform			Pair
Methods	20%	40%	60%	20%	40%
GCN	74.8 (0.9)	61.5 (0.3)	36.1 (1.8)	69.4 (0.6)	53.6 (0.9)
GAT	76.4 (0.7)	61.6 (4.4)	36.2 (2.9)	72.1 (0.7)	56.6 (2.5)
DGI	82.3 (0.2)	70.2 (0.2)	46.9 (0.2)	78.2 (0.2)	65.3 (0.3)
GCA	78.9 (0.7)	76.7 (0.7)	50.4 (3.7)	78.3 (1.0)	68.9 (1.7)
SUGRL	83.2 (0.4)	70.5 (0.1)	43.1 (0.2)	78.0 (0.2)	64.3 (0.6)

TABLE I : Classification accuracy (%) of all methods under different noise rates on four datasets. Best results are in bold.

Uniform noise (60%)
C1	C2	C3	Cora	Citeseer	DBLP	Computers
✓			59.9 (3.1)	36.0 (11.3)	71.5 (1.5)	79.3 (0.4)
	✓		62.6 (1.7)	39.9 (9.3)	74.1 (1.4)	79.4 (0.4)
		✓	70.8 (1.0)	64.0 (6.2)	80.3 (0.4)	80.7 (0.3)
✓	✓		63.1 (1.9)	38.4 (12.7)	73.6 (1.5)	79.6 (0.4)
✓		✓	70.7 (1.7)	64.6 (4.2)	80.3 (0.4)	80.8 (0.3)

TABLE II : Classification accuracy (%) of PaSta with different component combinations under the highest noise rates for uniform and pair noise.

Panel of Figure 5 — Fig. 3 : Node classification performance of PaSta with different annotators under the highest noise rates for uniform and pair noise.

Panel of Figure 6 — Fig. 4 : Sensitivity Analysis of the Parameter $\tau$ in PaSta .

Fig. 5 : Visualization of consensus partial label and one-hot pseudo-label (ten sampled noisy nodes).

Methods	Uniform			Pair
Methods	20%	40%	60%	20%	40%
GCN	89.90 (0.47)	87.03 (1.21)	81.07 (2.13)	85.74 (0.95)	64.27 (1.66)
GAT	89.86 (1.24)	88.17 (1.03)	81.24 (1.57)	87.36 (1.25)	68.67 (1.35)
DGI	90.22 (0.29)	85.20 (0.14)	83.49 (0.83)	85.97 (0.10)	63.43 (0.90)
GCA	86.91 (0.19)	82.94 (1.04)	73.73 (1.76)	86.03 (1.46)	77.34 (0.33)
SUGRL	91.47 (0.10)	89.22 (0.20)	83.25 (0.86)	89.92 (0.13)	70.88 (0.71)

TABLE III : Classification accuracy (%) of all methods under different noise rates on dataset Photo. Best results are in bold.

Panel of Figure 9 — Fig. 6 : Sensitivity Analysis of the Parameter $\lambda$ in PaSta .

Node Classification

Node classification assigns labels to graph entities by combining their features with information from connected nodes and graph structure.

See all 96 papers

Mar 14, 2026

Robust Self-Training with Closed-loop Label Correction for Learning from Noisy Labels

Training deep neural networks with noisy labels remains a significant challenge, often leading to degraded performance. Existing methods for handling label noise typically rely on either transition matrix, noise detection, or meta-learning techniques, but they often exhibit low utilization efficiency of noisy samples and incur high computational costs. In this paper, we propose a self-training label correction framework using decoupled bilevel optimization, where a classifier and neural correction function co-evolve. Leveraging a small clean dataset, our method employs noisy posterior simulation and intermediate features to transfer ground-truth knowledge, forming a closed-loop feedback system that prevents error amplification. Theoretical guarantees underpin the stability of our approach, and extensive experiments on benchmark datasets like CIFAR and Clothing1M confirm state-of-the-art performance with reduced training time, highlighting its practical applicability for learning from noisy labels.

Zhanhui Lin, Yanlin Liu, Sanping Zhou

Noisy-Label Learning Learning with Noisy Labels

Ask PDF

Nov 25, 2025

Pre-train to Gain: Robust Learning Without Clean Labels

Training deep networks with noisy labels leads to poor generalization and degraded accuracy due to overfitting to label noise. Existing approaches for learning with noisy labels often rely on the availability of a clean subset of data. By pre-training a feature extractor on the target dataset without labels using in-domain self-supervised learning (SSL), followed by standard supervised training on the same noisy dataset, we can train a more noise robust model without requiring a subset with clean labels. We evaluate both contrastive and non-contrastive SSL pre-training methods across datasets with synthetic and real-world label noise, demonstrating the broad applicability of our approach across large-scale datasets, diverse downstream tasks, and model architectures. Across all noise rates, in-domain self-supervised pre-training consistently improves classification accuracy and downstream label-error detection (F1 and Balanced Accuracy) compared with supervised training from scratch. The performance gap widens as the noise rate increases, demonstrating improved robustness. Notably, our approach achieves comparable results to ImageNet and DinoV2 pre-trained models at low noise levels, while substantially outperforming them under high noise conditions.

David Szczecina, Nicholas Pellegrino, Paul Fieguth

Systems Design Engineering University of Waterloo Waterloo, Ontario, Canada

Self-Supervised Pre-Training Learning with Noisy Labels

Ask PDF

Apr 10, 2026

Feature-Label Modal Alignment for Robust Partial Multi-Label Learning

In partial multi-label learning (PML), each instance is associated with a set of candidate labels containing both ground-truth and noisy labels. The presence of noisy labels disrupts the correspondence between features and labels, degrading classification performance. To address this challenge, we propose a novel PML method based on feature-label modal alignment (PML-MA), which treats features and labels as two complementary modalities and restores their consistency through systematic alignment. Specifically, PML-MA first employs low-rank orthogonal decomposition to generate pseudo-labels that approximate the true label distribution by filtering noisy labels. It then aligns features and pseudo-labels through both global projection into a common subspace and local preservation of neighborhood structures. Finally, a multi-peak class prototype learning mechanism leverages the multi-label nature where instances simultaneously belong to multiple categories, using pseudo-labels as soft membership weights to enhance discriminability. By integrating modal alignment with prototype-guided refinement, PML-MA ensures pseudo-labels better reflect the true distribution while maintaining robustness against label noise. Extensive experiments on both real-world and synthetic datasets demonstrate that PML-MA significantly outperforms state-of-the-art methods, achieving superior classification accuracy and noise robustness.

Yu Chen, Weijun Lv, Yue Huang +4

School of Automation, Guangdong University of Technology, Guangzhou 510006, China · Shenzhen Key Laboratory of Visual Object Detection and Recognition, Harbin Institute of Technology, Shenzhen 518055, China · School of Computer Science and Engineering, Sun Yat-Sen University, Guangzhou 510275, China

Multi-Label Classification Learning with Noisy Labels

Ask PDF

Abstract

Node Classification

Explore similar work

Robust Self-Training with Closed-loop Label Correction for Learning from Noisy Labels

Pre-train to Gain: Robust Learning Without Clean Labels

Feature-Label Modal Alignment for Robust Partial Multi-Label Learning