Scholarlink

<p>Attributed graph clustering is fundamental for understanding complex network structures, gaining significant research attention in recent years. The nonnegative matrix factorization (NMF) model has emerged as a powerful tool in this domain, excelling in pattern extraction and offering interpretable results. However, existing NMF-based methods often overlook the rich information contained in the attribute network’s manifold structure, focusing solely on topological network manifolds. To bridge this gap, we introduce Dual Orthogonal Graph Regularized Nonnegative Matrix Factorization for Attribute Graph Clustering (DOGNMF-AGC). This novel approach integrates a dual-graph model with NMF, enabling concurrent analysis of both topological and attributed network manifold structures. We enhance the model’s discriminative power through orthogonal constraints, leading to more distinct community delineations. Our contribution includes the development of iterative update optimization strategies for DOGNMF-AGC, ensuring guaranteed theoretical convergence. Extensive experiments conducted on nine diverse real-world datasets demonstrate DOGNMF-AGC’s superior clustering performance compared to state-of-the-art methods in attributed graph clustering.</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

A new method for attributed graph clustering with dual-manifold orthogonal matrix learning

  • 期刊论文

  • 发表日期：

    2025年10月13日

查看全文
  • TianYi Yu,
  • Shayan Nejadshamsi

摘要

Attributed graph clustering is fundamental for understanding complex network structures, gaining significant research attention in recent years. The nonnegative matrix factorization (NMF) model has emerged as a powerful tool in this domain, excelling in pattern extraction and offering interpretable results. However, existing NMF-based methods often overlook the rich information contained in the attribute network’s manifold structure, focusing solely on topological network manifolds. To bridge this gap, we introduce Dual Orthogonal Graph Regularized Nonnegative Matrix Factorization for Attribute Graph Clustering (DOGNMF-AGC). This novel approach integrates a dual-graph model with NMF, enabling concurrent analysis of both topological and attributed network manifold structures. We enhance the model’s discriminative power through orthogonal constraints, leading to more distinct community delineations. Our contribution includes the development of iterative update optimization strategies for DOGNMF-AGC, ensuring guaranteed theoretical convergence. Extensive experiments conducted on nine diverse real-world datasets demonstrate DOGNMF-AGC’s superior clustering performance compared to state-of-the-art methods in attributed graph clustering.