Simultaneously learning representation tensor and orthogonal projection for multi-view subspace clustering

Jinghao Li

doi:10.1016/j.neucom.2026.134255

Simultaneously learning representation tensor and orthogonal projection for multi-view subspace clustering

2026-06-11

0

PRE

AI

DOI：10.1016/j.neucom.2026.134255

Original

Original request for help

Save

Abstract

En 中文

Current tensor-learning approaches to multi-view subspace clustering operate directly on original data, where redundancy and noise may undermine the quality of the resulting clusters. To address the above problems, we design a new model, Simultaneously Learning Representation Tensor and Orthogonal Projection (SLRTOP) for multi-view subspace clustering. By introducing orthogonal projection learning, the model uses a projection matrix that maps the original data into a lower-dimensional space, where a representation tensor is constructed to alleviate the effects of redundancy and noise. Furthermore, SLRTOP jointly learns the representation matrix and the affinity matrix within a unified framework, leading to a more informative affinity structure. Experimental results on six multi-view datasets show that SLRTOP achieves superior clustering performance compared with several widely used existing methods.

Simultaneously learning representation tensor and orthogonal projection for multi-view subspace clustering

Abstract

Journal

Organization