pipette
ESEspañol

Total variation error bounds for Schr\"odinger-F\"ollmer samplers

Xiaojie Wang, Xiaoyan Zhang

Preprint

In the authors' words

Sampling from high-dimensional, multimodal target distributions plays a fundamental role in scientific computing, statistics and machine learning. Recently, the Schr\"odinger-F\"ollmer sampler (SFS) has been proposed by Huang et al. (IEEE Trans. Inform. Theory, 2025) as an efficient sampling algorithm that transports the degenerate distribution at to the target distribution at , without relying on long-time ergodicity. In this paper, we establish non-asymptotic error bounds for SFS in the total variation distance by deriving regularization estimates for the associated Kolmogorov equation via the Bismut-Elworthy-Li formula. When the drift can be computed exactly, a convergence rate of order was obtained under certain smoothness conditions on the drift with a uniform time step-size . When the drift is approximated by Monte Carlo methods, we obtain a convergence rate of order , where is the number of samples used in the Monte Carlo approximation. This yields a sharper dependence on than the existing bound in the -Wasserstein distance in the aforementioned paper, by turning the total variation error to the weak error analysis in the setting of non-smooth test functions. Numerical experiments are presented to illustrate the convergence rate with respect to the time step-size and the Monte Carlo sample size and the effectiveness of the proposed algorithm.

Main resultLimitation the authors admit

Appeared: Tuesday, November 10. arXiv. Preprint, not yet peer-reviewed.