Scholarlink

<p>Multi-label classification, which allows the assignment of multiple labels to an instance, is a key yet challenging problem in supervised learning. In this work, we propose a novel hybrid approach that combines the strengths of the extreme learning machine (ELM) and the twin support vector machine (TSVM) to address the challenges of robustness and scalability in multi-label classification, particularly in settings where deep learning is not practical due to limited training instances. In the first stage, ELM efficiently captures nonlinear patterns and reduces data complexity by transforming the original input space. Subsequently, a TSVM is trained in the transformed space to enhance generalization and mitigate the possibility of overfitting issue of ELM’s analytical solution. We introduce variants of the twin extreme learning machine that incorporate different loss functions, viz. <i>hinge loss</i>, <i>least squares loss</i>, and <i>weighted linear loss</i>. We evaluate the performance of the proposed approach on twelve benchmark multi-label datasets. The results demonstrate its superior generalization ability compared to the underlying baselines.</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-label classification with extreme learning machine and twin support vector machine: a novel hybrid framework

  • 期刊论文

  • 发表日期：

    2026年8月5日

查看全文
  • Amisha Bharti,
  • Vasudha Bhatnagar,
  • Vikas Kumar

摘要

Multi-label classification, which allows the assignment of multiple labels to an instance, is a key yet challenging problem in supervised learning. In this work, we propose a novel hybrid approach that combines the strengths of the extreme learning machine (ELM) and the twin support vector machine (TSVM) to address the challenges of robustness and scalability in multi-label classification, particularly in settings where deep learning is not practical due to limited training instances. In the first stage, ELM efficiently captures nonlinear patterns and reduces data complexity by transforming the original input space. Subsequently, a TSVM is trained in the transformed space to enhance generalization and mitigate the possibility of overfitting issue of ELM’s analytical solution. We introduce variants of the twin extreme learning machine that incorporate different loss functions, viz. hinge loss, least squares loss, and weighted linear loss. We evaluate the performance of the proposed approach on twelve benchmark multi-label datasets. The results demonstrate its superior generalization ability compared to the underlying baselines.