Scholarlink

<p>Gene selection from microarray data is a complex and important problem, particularly due to the high dimensionality and limited availability of labeled samples. In our previously published work, we proposed a semi-supervised method named Adaptive manifold learning for gene selection (AGMLGS), which demonstrated promising performance by using manifold learning on graph structures. Building on this foundation, we now introduce Hypergraph-based adaptive manifold learning for gene selection (HAMLGS) an improved approach that captures higher-order relationships among samples using a hypergraph representation. HAMLGS provides a closed-form solution that converges in a few iterations. Extensive experiments conducted on fifteen benchmark microarray datasets show that HAMLGS consistently outperforms AGMLGS and five other state-of-the-art feature selection methods in terms of average precision and F1 score. This comparison highlights the critical role of hypergraph modeling in effectively capturing complex structures in gene expression data. The MATLAB code for the proposed model is accessible at the following URL <a href="https://github.com/ml-lab-sau/HAMLGS">https://github.com/ml-lab-sau/HAMLGS</a>.</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

Hypergraph-based adaptive manifold learning for gene selection

  • 期刊论文

  • 发表日期：

    2026年2月2日

查看全文
  • Mamta Bhattarai Lamsal,
  • Reshma Rastogi

摘要

Gene selection from microarray data is a complex and important problem, particularly due to the high dimensionality and limited availability of labeled samples. In our previously published work, we proposed a semi-supervised method named Adaptive manifold learning for gene selection (AGMLGS), which demonstrated promising performance by using manifold learning on graph structures. Building on this foundation, we now introduce Hypergraph-based adaptive manifold learning for gene selection (HAMLGS) an improved approach that captures higher-order relationships among samples using a hypergraph representation. HAMLGS provides a closed-form solution that converges in a few iterations. Extensive experiments conducted on fifteen benchmark microarray datasets show that HAMLGS consistently outperforms AGMLGS and five other state-of-the-art feature selection methods in terms of average precision and F1 score. This comparison highlights the critical role of hypergraph modeling in effectively capturing complex structures in gene expression data. The MATLAB code for the proposed model is accessible at the following URL https://github.com/ml-lab-sau/HAMLGS.