Weighted Laplacian Flow: A Deterministic Particle Flow with Provable Convergence

DOI10.48550/arXiv.2608.21831MaRDI QIDQ7285026arXivFDO

Authors Weiye Gan, Tangjun Wang, Zuoqiang Shi

Publication date 22 August 2026

Published in arXiv

Tldr We introduce a deterministic particle-flow method that evolves samples from an initial distribution to a target density by solving a Poisson equation associated with the logarithmic density ratio between the two distributions. We prove global well-posedness and convergence in both L∞ distance and Kullback-Leibler divergence, achieving exact convergence under a sublinear forcing condition.

Full work available at URL https://doi.org/10.48550/arXiv.2608.21831

arXiv classification

math.APmath.NA

Cites work

• \{Euclidean, metric, and Wasserstein\} gradient flows: an overview
• A stochastic version of Stein variational gradient descent for efficient sampling
• Bayesian data analysis.
• Elliptic partial differential equations of second order
• Exponential Contraction in Wasserstein Distances for Diffusion Semigroups with Negative Curvature
• Exponential convergence of Langevin distributions and their discrete approximations
• High-dimensional Bayesian inference via the unadjusted Langevin algorithm
• Nonasymptotic convergence analysis for the unadjusted Langevin algorithm
• On the geometry of Stein variational gradient descent
• Regularized Stein Variational Gradient Flow
• Scaling Limit of the Stein Variational Gradient Descent: The Mean Field Regime
• SVGD as a kernelized Wasserstein gradient flow of the chi-squared divergence
• The deep Ritz method: a deep learning-based numerical algorithm for solving variational problems
• The Green function for uniformly elliptic equations
• The Variational Formulation of the Fokker--Planck Equation

This page was built for publication: Weighted Laplacian Flow: A Deterministic Particle Flow with Provable Convergence

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7285026)