Scholarlink

<p>Multi-view subspace clustering (MVSC) uncovers the intrinsic structure of multi-view datasets by integrating complementary information from diverse views. However, most existing methods neglect the simultaneous integration of consistency and complementarity in multi-view data, leading to degraded clustering performance. We propose a novel multi-view subspace clustering method, termed LRPMSC, which employs a projection matrix to map raw data into a lower-dimensional space. This mapping mitigates noise interference and preserves local structures while facilitating distance measurement between samples. Furthermore, LRPMSC adopts a matrix tri-factorization strategy to extract consistent structures from the representation matrix. The Schatten <i>p</i>-norm is imposed on the core matrix as a low-rank constraint, enhancing robustness against high-dimensional noise. Moreover, angular information and fusion mechanisms are incorporated to refine the final similarity matrix. To systematically evaluate LRPMSC, we conduct extensive experiments on nine benchmark datasets and compare its performance against twelve state-of-the-art multi-view clustering methods.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi-view subspace clustering via low-rank consensus projection matrix

  • 期刊论文

  • 发表日期：

    2026年1月28日

查看全文
  • Qing Zeng,
  • Yue Yu,
  • Wei Zhang,
  • Zizhu Fan

摘要

Multi-view subspace clustering (MVSC) uncovers the intrinsic structure of multi-view datasets by integrating complementary information from diverse views. However, most existing methods neglect the simultaneous integration of consistency and complementarity in multi-view data, leading to degraded clustering performance. We propose a novel multi-view subspace clustering method, termed LRPMSC, which employs a projection matrix to map raw data into a lower-dimensional space. This mapping mitigates noise interference and preserves local structures while facilitating distance measurement between samples. Furthermore, LRPMSC adopts a matrix tri-factorization strategy to extract consistent structures from the representation matrix. The Schatten p-norm is imposed on the core matrix as a low-rank constraint, enhancing robustness against high-dimensional noise. Moreover, angular information and fusion mechanisms are incorporated to refine the final similarity matrix. To systematically evaluate LRPMSC, we conduct extensive experiments on nine benchmark datasets and compare its performance against twelve state-of-the-art multi-view clustering methods.