Scholarlink

This paper introduces AFD-STA Net, a neural framework for predicting high-dimensional chaotic PDE systems by integrating adaptive filtering and spatiotemporal dynamics learning. The architecture combines adaptive exponential smoothing with position-aware decay, parallel attention for temporal–spatial dependencies, dynamic gated multiscale fusion, and deep projection networks. Experiments demonstrate robust accuracy across smooth and chaotic regimes, noise tolerance through adaptive filtering, and the critical role of spatiotemporal attention confirmed by ablation studies. Our code is available at https://anonymous.4open.science/r/AFD-STA-Net-B1E2 .

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

AFD-STA: Adaptive Filtering Denoising with Spatiotemporal Attention for Chaos Prediction

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Chunlin Gong,
  • Yin Wang,
  • Jingru Li,
  • Hanleran Zhang

摘要

This paper introduces AFD-STA Net, a neural framework for predicting high-dimensional chaotic PDE systems by integrating adaptive filtering and spatiotemporal dynamics learning. The architecture combines adaptive exponential smoothing with position-aware decay, parallel attention for temporal–spatial dependencies, dynamic gated multiscale fusion, and deep projection networks. Experiments demonstrate robust accuracy across smooth and chaotic regimes, noise tolerance through adaptive filtering, and the critical role of spatiotemporal attention confirmed by ablation studies. Our code is available at https://anonymous.4open.science/r/AFD-STA-Net-B1E2 .