Scholarlink

This paper explores the application of Liquid Neural Networks (LNN) to predict Worm humus production in agriculture. Specifically, we modify a simple Recurrent Neural Network (RNN) by including the dynamical behavior in the output calculation. We evaluate the efficacy of our approach by performing prediction tests on seven datasets collected from vermicomposting beds across three distinct agroclimatic zones in Valle del Cauca, Colombia, over eight months. Each dataset includes environmental variables recorded at 30 min intervals using an IoT-based sensor network. The proposed LNN model is compared to traditional deep learning approaches based on Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and 1D Convolutional Neural Networks (1D CNN). The LNN model reported an average Root Mean Squared Error (RMSE) percentage difference of 4.50% compared to the LSTM, GRU, and 1D CNN, demonstrating competitive performance.

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 Deep Learning Approach to Worm Humus Productivity Prediction: Application of Liquid Neural Networks

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jose L. Paniagua,
  • Velasco Juan Manuel Núñez,
  • Jesús A. López,
  • Fernando De la Prieta

摘要

This paper explores the application of Liquid Neural Networks (LNN) to predict Worm humus production in agriculture. Specifically, we modify a simple Recurrent Neural Network (RNN) by including the dynamical behavior in the output calculation. We evaluate the efficacy of our approach by performing prediction tests on seven datasets collected from vermicomposting beds across three distinct agroclimatic zones in Valle del Cauca, Colombia, over eight months. Each dataset includes environmental variables recorded at 30 min intervals using an IoT-based sensor network. The proposed LNN model is compared to traditional deep learning approaches based on Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and 1D Convolutional Neural Networks (1D CNN). The LNN model reported an average Root Mean Squared Error (RMSE) percentage difference of 4.50% compared to the LSTM, GRU, and 1D CNN, demonstrating competitive performance.