Scholarlink

<p>Incomplete multi-view clustering aims to address the key challenge of missing data in real-world multi-view scenarios. The methods learning consistent indicator has achieved promising results, among which Simultaneous Laplacian Embedding and Subspace Clustering (SLESC) is a recently proposed method. SLESC integrates similarity graph learning, Laplacian embedding, and discrete indicator matrix learning into a unified framework to directly output clustering results in one step. However, due to the missing data, this method adopts the strategy of adding zeros to obtain the complete representation matrix, which leads to inaccurate spectral embedding. In this paper, we propose Consistent Indicator Completion (CIC) to overcome this limitation. We learn the incomplete spectral embedding from the incomplete data and fill the spectral embedding of the missing data with consistent indicator. This strategy can avoid the propagation of incorrect information brought by adding zeros and improve computation speed by reducing the dimension of the representation matrix. In addition, we introduce the <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\( \ell _{2',1} \)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mi>ℓ</mi> <mrow> <msup> <mn>2</mn> <mo>′</mo> </msup> <mo>,</mo> <mn>1</mn> </mrow> </msub> </math></EquationSource> </InlineEquation>-norm to enhance view diversity. Experimental results on multiple real-world datasets validate the improvement of CIC on SLESC in terms of effectiveness and efficiency, and imply that CIC is a competitive incomplete multi-view clustering approach.</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

Incomplete multi-view clustering via consistent indicator completion

  • 期刊论文

  • 发表日期：

    2026年6月19日

查看全文
  • Chang Cheng,
  • Kewei Tang

摘要

Incomplete multi-view clustering aims to address the key challenge of missing data in real-world multi-view scenarios. The methods learning consistent indicator has achieved promising results, among which Simultaneous Laplacian Embedding and Subspace Clustering (SLESC) is a recently proposed method. SLESC integrates similarity graph learning, Laplacian embedding, and discrete indicator matrix learning into a unified framework to directly output clustering results in one step. However, due to the missing data, this method adopts the strategy of adding zeros to obtain the complete representation matrix, which leads to inaccurate spectral embedding. In this paper, we propose Consistent Indicator Completion (CIC) to overcome this limitation. We learn the incomplete spectral embedding from the incomplete data and fill the spectral embedding of the missing data with consistent indicator. This strategy can avoid the propagation of incorrect information brought by adding zeros and improve computation speed by reducing the dimension of the representation matrix. In addition, we introduce the \( \ell _{2',1} \) ℓ 2 ′ , 1 -norm to enhance view diversity. Experimental results on multiple real-world datasets validate the improvement of CIC on SLESC in terms of effectiveness and efficiency, and imply that CIC is a competitive incomplete multi-view clustering approach.