Structure-Aware Variational Learning of a Class of Generalized Diffusions

TL;DR

This paper introduces a structure-aware, energy-based learning framework for inferring potential functions in generalized diffusion processes, leveraging variational principles to improve robustness against noise and incomplete data.

Contribution

It develops a novel energy-dissipation based loss function grounded in the variational approach, avoiding PDE enforcement and enhancing robustness in learning stochastic dynamics.

Findings

The proposed method outperforms traditional regression approaches under noisy and incomplete observations.

Numerical experiments show improved robustness with respect to observation time and noise levels.

The energy-based loss effectively captures the underlying dynamics across multiple dimensions.

Abstract

Learning the underlying potential energy of stochastic gradient systems from partial and noisy observations is a fundamental problem arising in physics, chemistry, and data-driven modeling. Classical approaches often rely on direct regression of governing equations or velocity fields, which can be sensitive to noise and external perturbations and may fail when observations are incomplete. In this work, we propose a structure-aware, energy-based learning framework for inferring unknown potential functions in generalized diffusion processes, grounded in the energetic variational approach. Starting from the energy-dissipation law associated with the Fokker-Planck equation, we construct loss functions based on the De Giorgi dissipation functional, which consistently couple the free energy and the dissipation mechanism of the system. This formulation avoids explicit enforcement of the governing partial differential equation and preserves the underlying variational structure of the dynamics. Through numerical experiments in one, two, and three dimensions, we demonstrate that the proposed energy-based loss exhibits enhanced robustness with respect to observation time, noise level, and the diversity and amount of available training data. These results highlight the effectiveness of energy-dissipation principles as a reliable foundation for learning stochastic diffusion dynamics from data.