New hybrid framework learns physics ODEs with neural networks

Researchers have developed a novel hybrid framework that combines neural networks with physics-based ordinary differential equations (ODEs) to model dynamical systems. This method is particularly useful when some ODE components are unknown and only a subset of state variables are measurable. The approach alternates between estimating latent states using a Rauch--Tung--Striebel (RTS) smoother and then using these smoothed trajectories to train neural network parameters via backpropagation. Evaluations on benchmark systems demonstrate the framework's ability to learn missing ODE components from incomplete data, while also improving latent-state reconstruction and long-horizon prediction. AI

IMPACT This research could advance the modeling of complex dynamical systems where full state information is unavailable, potentially impacting fields like scientific simulation and robotics.

RANK_REASON The cluster contains an academic paper detailing a new method for learning physics-based neural differential models.

Read on arXiv cs.LG →

• alphaXiv
• artificial neural network
• arXiv
• CatalyzeX
• Connected Papers
• DagsHub
• Gotit.pub
• Hugging Face
• IArxiv
• Litmaps
• ordinary differential equations
• Rauch--Tung--Striebel
• ScienceCast
• scite Smart Citations

• paper
• other

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →