Safe-Subspace Pseudo-Label Refinement for Source-Free Graph Domain Adaptation

2026-05-30

AI Summary

This work addresses the unreliability of pseudo-labels on target graphs when source data is unavailable, a challenge exacerbated by feature and topological shifts. To mitigate this, the authors propose a selective pseudo-labeling approach that identifies a “safe subspace” characterized by semantic and structural consistency via constrained posterior discrepancy theory. Hard pseudo-label supervision is applied exclusively within this safe subspace, while noise-robust soft regularization is employed elsewhere. The selection mechanism integrates multiple criteria, including committee-based confidence from the source model, intrinsic structural representations derived from graph contrastive learning, and neighborhood consistency. Extensive experiments demonstrate that the method achieves robust and state-of-the-art source-free graph transfer performance across diverse image and real-world graph datasets under various domain shifts.

Abstract

Source-free graph domain adaptation (SF-GDA) aims to adapt source-trained graph models to unlabeled target graphs when source graphs are no longer accessible. A central obstacle is pseudo-label reliability: under feature and topological shifts, source-induced predictions may become confidently wrong, and indiscriminate self-training can amplify systematic errors through graph message passing. This paper studies SF-GDA from a selective pseudo-labeling perspective. Instead of assuming globally bounded pseudo-label noise over the entire target domain, we identify a confidence-consistent safe subspace on which pseudo-label noise can be controlled under restricted posterior discrepancy, and derive a target-risk decomposition that separates safe-subspace fitting error, selected-label noise, and uncertain-set risk. Guided by this analysis, we propose SafeSubspace Pseudo-Label Refinement (S$^2$PLR), a source-free graph adaptation framework that applies hard pseudo-label supervision only to target graphs supported by both semantic and structural evidence. Specifically, S$^2$PLR estimates semantic reliability using source-committee confidence and disagreement, learns a targetintrinsic structural representation via graph contrastive learning, verifies pseudo-labels through neighborhood consistency, and exploits the remaining uncertain samples with noise-tolerant soft regularization rather than unreliable hard labels. Experiments on image and real-world graph benchmarks under different domain shifts demonstrate that S$^2$PLR achieves robust and competitive performance across diverse source-free transfer settings.

Problem

• source-free graph domain adaptation
• pseudo-label reliability
• feature and topological shifts
• graph message passing
• domain adaptation

Innovation

• safe subspace
• pseudo-label refinement
• source-free graph domain adaptation
• graph contrastive learning
• neighborhood consistency