Data-Driven Free-Energy Learning from Trajectories via Structure-Preserving Exponential Time Differencing

DOI10.48550/arXiv.2610.11931MaRDI QIDQ7374664arXivFDO

Authors Wenshuai Hu, Guanghua Ji, Xiao Li

Publication date 8 October 2026

Published in arXiv

Tldr We learn a structure-preserving evolution operator for gradient flows from trajectories using a novel Duhamel-based learning framework that reconstructs an inferred free energy, enabling unconditional dissipation of the learned energy and maximum-bound principle preservation.

Full work available at URL https://doi.org/10.48550/arXiv.2610.11931

Mathematics Subject Classification ID

Reaction-diffusion equations (35K57) Finite difference methods for initial value and initial-boundary value problems involving PDEs (65M06) Stability and convergence of numerical methods for initial value and initial-boundary value problems involving PDEs (65M12) Error bounds for initial value and initial-boundary value problems involving PDEs (65M15)

arXiv classification

math.NA

Cites work

• An adaptive energy-based sequential method for training PINNs to solve gradient flow equations
• Energy Dissipation Preserving Physics Informed Neural Network for Allen-Cahn Equations
• Energy-dissipative evolutionary deep operator neural networks
• Fourier Neural Operator for Parametric Partial Differential Equations
• Free energy of a nonuniform system. I: Interfacial free energy
• Machine learning materials physics: integrable deep neural networks enable scale bridging by learning free energy functions
• Orientability and energy minimization in liquid crystal models
• Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations
• Physics-Informed Neural Operator for Learning Partial Differential Equations
• Variational Onsager Neural Networks (VONNs): A thermodynamics-based variational learning strategy for non-equilibrium PDEs

This page was built for publication: Data-Driven Free-Energy Learning from Trajectories via Structure-Preserving Exponential Time Differencing

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7374664)