PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
May 13, 2026Concurrency and Computation Practice and Experience0 citations

Efficient Self‐Guided One‐Pass Multi‐View Subspace Clustering

View Full Paper
HDHaoyu DuanSLShigang LiuLWLeqi Wu

Key Points

  • This research aims to enhance clustering performance by addressing limitations of existing multi-view clustering methods.
  • Proposed an Efficient Self-guided One-pass Multi-view Subspace Clustering (ESOMSC) method.
  • Combined view-shared anchor learning and global-guided-local self-guidance learning.
  • Enforced a rank constraint on the Laplacian matrix of the similarity matrix.
  • The method directly generates clustering labels without post-processing steps.
  • Achieved linear time complexity relative to the sample size.
  • Demonstrated improved effectiveness compared to state-of-the-art clustering methods.

Abstract

ABSTRACT In recent years, many multi‐view subspace clustering methods have been proposed to enhance the performance of clustering. But most of the existing methods still face several significant issues. First, these methods are unable to simultaneously capture consistency and complementary information. Second, these methods divide data into groups by a data similarity matrix. Therefore, the clustering result relies heavily on the similarity matrix. Because similarity estimation and clustering are usually done separately, the estimated similarity is not necessarily the most suitable for clustering and may reduce clustering performance. To address these issues, we propose an Efficient Self‐guided One‐pass Multi‐view Subspace Clustering (ESOMSC) method. Specifically, we combine the view‐shared anchor learning and global‐guided‐local self‐guidance learning into a uniform framework to improve clustering performance. And the framework can balance consistency and diversity across multiple views. At the same time, a rank constraint is enforced on the Laplacian matrix of the similarity matrix, ensuring that the number of connected components in the resulting graph precisely matches the number of clusters. So our method directly generates clustering labels, eliminating the need for post‐processing steps. Moreover, we prove that the proposed ESOMSC achieves linear time complexity relative to the sample number. Extensive experiments on real datasets demonstrate the effectiveness and efficiency of our proposed method compared to the state‐of‐the‐art clustering methods.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Duan et al. (2026) studied this question.

synapsesocial.com/papers/6a04153d79e20c90b4445147https://doi.org/10.1002/cpe.70731
Ask AI
Helpful
Bookmark
Share
View Full Paper