Scholarlink

<p>This paper introduces a novel <i>stronger than</i> relation among irreducible testors and formally defines the concept of <i>strongest irreducible testors</i>. A physics-inspired algorithm—based on Newton’s law of gravitation—is proposed for their computation. To the best of our knowledge, this is the first application of a physics-inspired optimization algorithm for computing testors. Additionally, the concept of <i>core</i> from rough set theory is incorporated to improve the efficiency of the search process. The effectiveness of strongest irreducible testors for data dimensionality reduction is demonstrated through their application in supervised classification tasks, showing promising results. On average, the strongest testors achieved a 60% reduction in the representation space. Moreover, in experiments with four classifiers, the reduced representations yielded classification results that were never worse than those obtained with the original data.</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

On the Strongest Irreducible Testors and Their Computation Using a Physics-Inspired Algorithm

  • 期刊论文

  • 发表日期：

    2025年7月15日

查看全文
  • Manuel S. Lazo-Cortés,
  • Alejandro Rodríguez-Molina,
  • Sandra S. Roblero-Aguilar,
  • Rodolfo Velázquez-Mancilla

摘要

This paper introduces a novel stronger than relation among irreducible testors and formally defines the concept of strongest irreducible testors. A physics-inspired algorithm—based on Newton’s law of gravitation—is proposed for their computation. To the best of our knowledge, this is the first application of a physics-inspired optimization algorithm for computing testors. Additionally, the concept of core from rough set theory is incorporated to improve the efficiency of the search process. The effectiveness of strongest irreducible testors for data dimensionality reduction is demonstrated through their application in supervised classification tasks, showing promising results. On average, the strongest testors achieved a 60% reduction in the representation space. Moreover, in experiments with four classifiers, the reduced representations yielded classification results that were never worse than those obtained with the original data.