Research map / Probabilistic and scientific modeling / Dynamical systems
Frequency-Structured Hamiltonian Neural Network for Multi-Timescale Dynamics
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes FS-HNN, a Hamiltonian neural network that decomposes the Hamiltonian into learned frequency-structured components to model stiff multi-timescale coupled fast-slow dynamics while preserving structure.
Problem
Scalar-Hamiltonian neural networks inherit the spectral bias of deep networks, which limits learning of stiff multi-timescale systems with coupled fast and slow dynamics.
Approach
Decomposes the system Hamiltonian into learned frequency-structured components instead of a single scalar Hamiltonian parameterization, so that fast and slow timescales are represented separately while conservation structure is kept.
Experimental setup
Builds on: Hamiltonian Neural Networks
Closest papers by meaning
- Hamiltonian Neural PDE Solvers through Functional Approximation (NeurIPS 2025 · similarity 0.90)
- Stable Port-Hamiltonian Neural Networks (NeurIPS 2025 · similarity 0.88)
- Physics-Informed Regularization for Domain-Agnostic Dynamical System Modeling (NeurIPS 2024 · similarity 0.87)
- Stable Long-Horizon PDE Forecasting via Latent Structured Spectral Propagators (NeurIPS 2026 · similarity 0.86)
- Discovering Phase Space Structure in Learned Hamiltonian Systems (NeurIPS 2026 · similarity 0.86)
- Hamiltonian Score Matching and Generative Flows (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.