Scholarlink

<p>As a support vector-based clustering approach, the twin support vector clustering (TWSVC) model has achieved great success. Equipped with a kernel function, the kernel-based TWSVC model is able to handle nonlinear clustering tasks. However, it takes much extra effort selecting a proper kernel function and tuning the hyper-parameters. In this research, a novel kernel-free nonlinear quadratic least squares TWSVC model is proposed. Without utilizing any kernel functions, it captures the clusters by directly producing quadratic surfaces. By employing the idea of TWSVC, each cluster is characterized individually, which brings a better generality of the model. In addition, the proposed model adopts the least squares constraints so that the clustering results can be efficiently obtained by solving a series of linear systems of equations. Computational experiments on artificial and public benchmark datasets are conducted to validate the clustering performance of the proposed model. Moreover, the proposed model is applied to the steel 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

Kernel-free quadratic least squares twin support vector clustering

  • 期刊论文

  • 发表日期：

    2025年4月29日

查看全文
  • Zheming Gao,
  • Haojie Fu,
  • Min Huang,
  • Jian Luo

摘要

As a support vector-based clustering approach, the twin support vector clustering (TWSVC) model has achieved great success. Equipped with a kernel function, the kernel-based TWSVC model is able to handle nonlinear clustering tasks. However, it takes much extra effort selecting a proper kernel function and tuning the hyper-parameters. In this research, a novel kernel-free nonlinear quadratic least squares TWSVC model is proposed. Without utilizing any kernel functions, it captures the clusters by directly producing quadratic surfaces. By employing the idea of TWSVC, each cluster is characterized individually, which brings a better generality of the model. In addition, the proposed model adopts the least squares constraints so that the clustering results can be efficiently obtained by solving a series of linear systems of equations. Computational experiments on artificial and public benchmark datasets are conducted to validate the clustering performance of the proposed model. Moreover, the proposed model is applied to the steel clustering.