Scholarlink

<p>In this paper, we propose a novel binary classifier based on a modified Lengyel–Epstein (LE) equation. By introducing an additional modulatory term, the modified LE system progressively forms a stable nonlinear classification boundary. The equation is numerically solved using the explicit Euler method, chosen for its computational efficiency. To assess robustness, we conduct experiments on the two-moon dataset. Results demonstrate that the classification boundary produced by our model converges to a stable state with low parameter sensitivity and strong stability. Furthermore, to validate effectiveness, we test the classifier on Archimedean spiral data under three different configurations. Despite the high density and close proximity of data points, the classifier consistently achieves an accuracy exceeding 98.5%, underscoring its superiority. Finally, we apply our method to real-world medical diagnostic tasks, achieving 100% accuracy on EEG signal data and small-sample MRIs datasets. On large-sample MRIs classification tasks, our approach outperforms all competing algorithms, highlighting its strong generalization ability.</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 novel binary data classifier based on a modified Lengyel-Epstein equation

  • 期刊论文

  • 发表日期：

    2026年3月17日

查看全文
  • Haozhe Wang,
  • Zean Dai,
  • Haixiao Wang,
  • Jian Wang

摘要

In this paper, we propose a novel binary classifier based on a modified Lengyel–Epstein (LE) equation. By introducing an additional modulatory term, the modified LE system progressively forms a stable nonlinear classification boundary. The equation is numerically solved using the explicit Euler method, chosen for its computational efficiency. To assess robustness, we conduct experiments on the two-moon dataset. Results demonstrate that the classification boundary produced by our model converges to a stable state with low parameter sensitivity and strong stability. Furthermore, to validate effectiveness, we test the classifier on Archimedean spiral data under three different configurations. Despite the high density and close proximity of data points, the classifier consistently achieves an accuracy exceeding 98.5%, underscoring its superiority. Finally, we apply our method to real-world medical diagnostic tasks, achieving 100% accuracy on EEG signal data and small-sample MRIs datasets. On large-sample MRIs classification tasks, our approach outperforms all competing algorithms, highlighting its strong generalization ability.