Scholarlink

<p>Deep Learning (DL) has been applied, over the last years, to a variety of tasks, from image classification to text generation. With the increase of the model parameters, a major issue is still their interpretability. In this context, Neuro-Fuzzy (NF) systems appear as architectures that are similar to Neural Networks (NN), but also presenting intrinsic interpretable parameters. In this work, <i>ANFISpy</i> is introduced as a novel <i>Python</i> package for the implementation of Neuro-Fuzzy models, based on the <i>PyTorch</i> framework. The implementation is discussed in detail, as well as the main visualization tools. <i>ANFISpy</i> was compared to other <i>Python</i> packages, showing comparable performance. Finally, real data applications are presented, including a time-series forecasting problem, exemplifying the flexibility of the proposed implementation for the creation of new NF architectures.</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

ANFISpy: a python package for neuro-fuzzy models

  • 期刊论文

  • 发表日期：

    2026年2月19日

查看全文
  • Matheus Zaia Monteiro,
  • Vinicius Francisco Wasques

摘要

Deep Learning (DL) has been applied, over the last years, to a variety of tasks, from image classification to text generation. With the increase of the model parameters, a major issue is still their interpretability. In this context, Neuro-Fuzzy (NF) systems appear as architectures that are similar to Neural Networks (NN), but also presenting intrinsic interpretable parameters. In this work, ANFISpy is introduced as a novel Python package for the implementation of Neuro-Fuzzy models, based on the PyTorch framework. The implementation is discussed in detail, as well as the main visualization tools. ANFISpy was compared to other Python packages, showing comparable performance. Finally, real data applications are presented, including a time-series forecasting problem, exemplifying the flexibility of the proposed implementation for the creation of new NF architectures.