.png){kind=icon}
.png){kind=icon}
Return
Cross-view structure awareness network for attribute-missing graph clustering
.png){kind=icon}
PRE
AI
DOI：10.1016/j.neucom.2026.134067.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Attribute-missing graph clustering, a fundamental yet challenging task in data analysis, endeavors to divide nodes into distinct groups when some node attributes are missing. Recently, researchers have proposed multi-view graph representation learning to perform data imputation, leading to significant progress in attribute-missing graph clustering. However, prevalent multi-view methods, which process each view independently, fail to exploit cross-view structural correlations for attribute-missing nodes. To bridge this gap, this paper proposes the ross-view tructure awareness network for ttribute-missing raph lustering (CSAGC). Essentially, CSAGC generates fine-grained local structural patterns, which guide network optimization from both explicit and implicit perspectives. Specifically, we first generate multiple input views and subsequently apply a granular-ball sampling strategy to partition each view’s attribute matrix into an adaptive number of granular balls, which serve as local structural priors for explicit cross-view structure awareness. Meanwhile, we design an implicit topology supervision mechanism that integrates constructed topologies based on local structural patterns as self-supervised signals, thus implicitly guiding the network to learn more stable and clustering-friendly graph representation. Finally, a novel multi-view learning method is proposed to maximize the consistency across local and global views. Extensive experiments demonstrate the superiority of CSAGC against state-of-the-art competitors.
Keywords:
attribute-missing graph clustering
multi-view graph representation learning
cross-view structural correlation
granular-ball sampling
self-supervised topology supervision
Journal
IF:
6.5
Papers:
2.5W
Citations:
6.5W