New neural network predicts Hamiltonian chaos beyond training data

Researchers have developed a new type of Hamiltonian neural network (RF-HNN) capable of predicting Hamiltonian chaos in dynamical systems, even in regimes not present in its training data. Unlike conventional HNNs, the RF-HNN can extrapolate from predominantly regular dynamics to predict the emergence and growth of chaotic regions. This advancement was demonstrated across four Hamiltonian systems, including the Hénon-Heiles system, showing the network's ability to reproduce complex dynamical behaviors like the breakup of regular structures. AI

IMPACT This research could enable more accurate long-term predictions in complex dynamical systems, impacting fields from physics to engineering.

RANK_REASON This is a research paper detailing a new method for predicting Hamiltonian chaos using a novel neural network architecture. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Hamiltonian neural networks
• Hénon-Heiles system
• Hugging Face
• random-feature Hamiltonian neural networks
• RF-HNN

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →