Variational Optimality of Föllmer Processes in Generative Diffusions

arXiv:2602.10989v11 citationsh-index: 10
Originality Incremental advance
AI Analysis

This provides a new variational characterization of Föllmer processes for researchers in generative modeling and diffusion processes, offering incremental theoretical insights.

The paper tackles the problem of constructing generative diffusions to transport a point mass to a target distribution, proving that tuning the diffusion coefficient to minimize estimation error selects a Föllmer process in closed form, which makes the path-space Kullback-Leibler divergence independent of the interpolation schedule.

We construct and analyze generative diffusions that transport a point mass to a prescribed target distribution over a finite time horizon using the stochastic interpolant framework. The drift is expressed as a conditional expectation that can be estimated from independent samples without simulating stochastic processes. We show that the diffusion coefficient can be tuned \emph{a~posteriori} without changing the time-marginal distributions. Among all such tunings, we prove that minimizing the impact of estimation error on the path-space Kullback--Leibler divergence selects, in closed form, a Föllmer process -- a diffusion whose path measure minimizes relative entropy with respect to a reference process determined by the interpolation schedules alone. This yields a new variational characterization of Föllmer processes, complementing classical formulations via Schrödinger bridges and stochastic control. We further establish that, under this optimal diffusion coefficient, the path-space Kullback--Leibler divergence becomes independent of the interpolation schedule, rendering different schedules statistically equivalent in this variational sense.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes