Fast multi-view clustering via anchor label transmit with tensor structure constraint

Huimin Chen, Runxin Zhang, Yu Duan, Rong Wang, Feiping Nie

Research output: Contribution to journalArticlepeer-review

Abstract

In recent years, multi-view learning has emerged as a pivotal area of research within machine learning. Data from multiple views typically share consistent structures and contain complementary information. Therefore, how to combine information from multiple views to learn consistent structures has become a central theme of these studies. In this paper, we propose a Fast Multi-view Clustering via Anchor Label Transmit with Tensor Structure Constraint (F-MALT) to achieve rapid and accurate learning from multi-view data. First, we approximate the data membership for each view by combining the clustering of a small set of anchors and the neighborhood structure within the anchor graph. Then, we construct a low-rank tensor to extract consistent structures across views to effectively exploit complementary information and resist noise and outliers in individual views. Such a consistent structure can further guide anchor clustering to obtain the best membership. Experiments on synthetic data confirmed the robustness of our F-MALT against noise, while experiments on benchmark datasets demonstrated its efficiency.

Original languageEnglish
Article number126878
JournalExpert Systems with Applications
Volume274
DOIs
StatePublished - 15 May 2025

Keywords

  • Fast learning of membership
  • Graph based clustering
  • Low-rank tensor learning
  • Multi-view clustering

Fingerprint

Dive into the research topics of 'Fast multi-view clustering via anchor label transmit with tensor structure constraint'. Together they form a unique fingerprint.

Cite this