Uniform-in-time Propagation-of-Chaos for Stein Variational Gradient Descent

2026-06-30

AI Summary

This work addresses the lack of long-time consistency between finite-particle Stein variational gradient descent (SVGD) systems and their mean-field limit by introducing a truncation strategy that integrates finite-time propagation-of-chaos estimates with long-time convergence analysis. The authors develop a finite-dimensional matrix kernel theory and, for the first time, establish logarithmic-in-time uniform propagation-of-chaos bounds for SVGD under both the Langevin kernel Stein discrepancy and the Wasserstein distance. In the Gaussian target setting, they obtain a parametrically uniform convergence rate of $N^{-1/2}$ in physical time. By leveraging the principle of diffeomorphic conjugacy, these results are further extended to nonlinear multimodal distributions, revealing a fundamental distinction between distributional metrics and finite-dimensional observables in their long-time behavior.

Abstract

We study uniform-in-time propagation-of-chaos for continuous-time Stein Variational Gradient Descent (SVGD). Classical finite-time propagation-of-chaos estimates for mean-field systems typically deteriorate rapidly with time and therefore do not directly explain the long-time relation between the finite-particle system and its mean-field limit. We obtain two complementary classes of uniform-in-time propagation-of-chaos results. For broad distributional metrics, we introduce a cutoff strategy which combines finite-time propagation-of-chaos estimates up to an $N$-dependent horizon with independent quantitative long-time convergence estimates for the finite-particle and mean-field SVGD flows. This yields uniform-in-averaging-time propagation-of-chaos bounds in Langevin kernel Stein discrepancy, Wasserstein-1 distance, and Wasserstein-2 distance, with logarithmic or iterated-logarithmic rates depending on the metric, target and kernel class. We also develop a finite-dimensional theory for matrix-valued finite-rank kernels. For Gaussian targets with bilinear kernels, the SVGD dynamics close exactly on first and second moments, yielding genuine uniform-in-physical-time parametric propagation-of-chaos rates in finite-dimensional Stein-feature metrics. We then prove a conjugacy principle showing that these feature-level estimates transfer to conjugate target-kernel pairs under orientation-preserving diffeomorphisms, thereby extending the theory to broad classes of nonlinear, including multimodal, targets. Together, these results highlight the contrast between generic distributional metrics, for which our general approach yields logarithmic rates, and closed finite-dimensional Stein observables, for which parametric $N^{-1/2}$ propagation-of-chaos rates persist uniformly in time.

Problem

• propagation-of-chaos
• Stein Variational Gradient Descent
• mean-field limit
• long-time behavior
• uniform-in-time

Innovation

• uniform-in-time propagation-of-chaos
• Stein Variational Gradient Descent
• finite-rank kernels
• moment closure
• conjugacy principle