Skip to main navigation Skip to search Skip to main content

USBD: Universal Structural Basis Distillation for Source-Free Graph Domain Adaptation

  • Yingxu Wang (Co-first Author)
  • , Kunyu Zhang (Co-first Author)
  • , Mengzhu Wang
  • , Siyang Gao
  • , Nan Yin*
  • *Corresponding author for this work

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

1 Downloads (CityUHK Scholars)

Abstract

Source-Free Graph Domain Adaptation (SF-GDA) enables privacypreserving knowledge transfer across graph datasets. Although recent works incorporate structural information, they implicitly condition adaptation on the smoothness priors of source-trained GNNs, limiting generalization to structurally distinct targets. This dependency is critical under significant topological shifts, where the source model misinterprets distinct topological patterns unseen in the source domain as noise, rendering pseudo-label-based adaptation unreliable. To overcome this limitation, we propose the Universal Structural Basis Distillation (USBD), a framework that shifts the paradigm from adapting a biased model to learning a universal structural basis for SF-GDA. Instead of adapting a biased source model to a specific target, we construct a structure-agnostic basis that proactively covers the full spectrum of potential topological patterns. Specifically, USBD employs a bi-level optimization framework to distill the source dataset into a compact structural basis. By enforcing the prototypes to span the full Dirichlet energy spectrum, the learned basis explicitly captures diverse topological motifs, ranging from low-frequency clusters to high-frequency chains, beyond those present in the source. This ensures that the learned basis creates a comprehensive structural covering capable of handling targets with disparate structures. For inference, we introduce a spectral-aware ensemble mechanism that dynamically activates the optimal prototype combination based on the spectral fingerprint of the target graph. Extensive experiments on benchmarks demonstrate that USBD significantly outperforms state-of-the-art methods, particularly in scenarios with severe structural shifts, while achieving superior computational efficiency by decoupling the adaptation cost from the target data scale. © 2026 Owner/Author.
Original languageEnglish
Title of host publicationKDD '26 - Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2
Place of PublicationNew York, NY
PublisherAssociation for Computing Machinery
Pages5125-5136
Number of pages12
ISBN (Print)9798400722592
DOIs
Publication statusPublished - Aug 2026
Event32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD 2026) - International Convention Center Jeju (ICC Jeju), Jeju Island, Korea, Republic of
Duration: 9 Aug 2026 → 13 Aug 2026
https://kdd2026.kdd.org/

Publication series

NameProceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
Volume2-B
ISSN (Print)2154-817X

Conference

Conference32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD 2026)
Abbreviated titleACM KDD 2026
PlaceKorea, Republic of
CityJeju Island
Period9/08/26 → 13/08/26
Internet address

Bibliographical note

Full text of this publication does not contain sufficient affiliation information. Related Research Unit(s) information for this record is supplemented by the author(s) concerned.

Research Keywords

  • domain adaptation
  • graph neural networks
  • transfer learning

Publisher's Copyright Statement

  • This full text is made available under CC-BY 4.0. https://creativecommons.org/licenses/by/4.0/

Fingerprint

Dive into the research topics of 'USBD: Universal Structural Basis Distillation for Source-Free Graph Domain Adaptation'. Together they form a unique fingerprint.

Cite this