Langevin-Informed Transfer Learning: Replacing Target Samples by Black-Box Feedback

2026-10-01

AI Summary

This study addresses the challenge of extracting low-dimensional slow dynamics in stochastic dynamical systems where target trajectories are inaccessible and only biased samples are available. To this end, we propose the LITL framework, which relies solely on black-box feedback. By leveraging Dirichlet representation learning, the method captures the spectral structure and projected drift of the target infinitesimal generator, while introducing spherical variants to optimize the guided control of normalized latent representations, thereby achieving Langevin dynamics reconstruction via spectral operator learning. Experimental results demonstrate that the proposed approach successfully recovers physical timescales and spherical symmetry from biased simulations, establishes a dynamical structure for generative models, and enables post-hoc guided control within the latent space.

Abstract

Many scientific and machine learning systems, from molecular dynamics to diffusion models and beyond, are governed by stochastic dynamics with low-dimensional structure, evolving on slow timescales. However, target trajectories, used to identify and interpret such dynamics, are often inaccessible: only biased or static samples that explore the underlying manifold are available. We introduce Langevin-Informed Transfer Learning (LITL), a framework for recovering target Langevin dynamics from biased source samples using only black-box feedback. LITL learns the leading spectral structure of the target infinitesimal generator and the projected drift through Dirichlet representation learning, enabling kinetic reconstruction in spectral form and slow-manifold gradient field estimation. We further introduce a spherical variant well suited to steering normalized latent representations commonly used in learning systems toward desired objectives. We establish finite-sample guarantees for eigenvalue, eigenfunction, and projected drift estimation in Sobolev norms, thereby ensuring generalization of these quantities and their first-order derivatives. Empirically, LITL recovers physical transition timescales from biased molecular simulations, builds kinetic structure from static samples of generative models, reconstructs spherical symmetries of physical systems, and enables post-hoc latent steering of trained neural networks under black-box feedback. Together, these results position spectral operator learning as a practical framework for recovering stochastic dynamics under distribution shift and unlock applications across machine learning and the physical sciences.

Problem

• Langevin dynamics
• transfer learning
• black-box feedback
• biased samples
• spectral operator learning

Innovation

• Langevin dynamics
• Transfer learning
• Spectral operator learning
• Black-box feedback
• Dirichlet representation learning