Scholarlink

Alzheimer’s Disease (AD) is a progressive neurological disorder with profound societal and healthcare impacts. Recent advancements in speech-based AD detection highlight Deep Learning (DL) methods’ potential to identify cognitive impairments through speech analysis. However, conventional DL models’ “black box” nature raises interpretability concerns in critical healthcare settings. This study introduces Kolmogorov–Arnold Networks (KANs) with learnable activation functions, enhancing both interpretability and efficiency. Using the ADReSSo dataset, KAN achieves an accuracy of \(82.81\%\) and a Cohen’s Kappa score of \(0.7893\) . Additionally, KAN’s transparent architecture aids AD speech understanding, balancing accuracy and interpretability for detection.

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

Kolmogorov–Arnold Networks for Speech-Based Alzheimer’s Detection

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Anass El Hallani,
  • Adil Chakhtouna,
  • Abdellah Adib

摘要

Alzheimer’s Disease (AD) is a progressive neurological disorder with profound societal and healthcare impacts. Recent advancements in speech-based AD detection highlight Deep Learning (DL) methods’ potential to identify cognitive impairments through speech analysis. However, conventional DL models’ “black box” nature raises interpretability concerns in critical healthcare settings. This study introduces Kolmogorov–Arnold Networks (KANs) with learnable activation functions, enhancing both interpretability and efficiency. Using the ADReSSo dataset, KAN achieves an accuracy of \(82.81\%\) and a Cohen’s Kappa score of \(0.7893\) . Additionally, KAN’s transparent architecture aids AD speech understanding, balancing accuracy and interpretability for detection.