Searcharxiv⌕ Search

arXiv · 2610.35892

Accelerating Langevin Dynamics with Weighted Laplacian Flow: Global Well-Posedness and Exponential Convergence

Abstract

We study an accelerated Langevin equation in which the classical overdamped Langevin drift is supplemented by a nonlocal gradient velocity determined through a target-weighted Poisson equation. The target $p$ is only assumed to be strictly positive and need not be normalized. Writing the dynamics in terms of $ω=\log(p/q)$ yields a coupled parabolic-elliptic system whose parabolic component contains the usual Langevin diffusion, whereas the elliptic component generates a global transport. The system is invariant under multiplication of $p$ by a positive constant. On a bounded smooth domain with homogeneous Neumann boundary conditions, we establish global existence and uniqueness of classical solutions for strictly positive $C^{2+α}$ target weights and compatible $C^{2+α}$ initial data. For long-time behavior, the oscillation of $ω$ decays at the explicit rate $e^{-t}$ and the solution density exponentially converges to target density. Moreover, an exact entropy-dissipation identity holds relative to the target density, in which the relative Fisher information and the Jeffreys divergence give the two dissipation mechanisms associated with the Langevin and weighted-Poisson components. In particular, KL divergence also converges to zero with rate $e^{-t}$ without a convexity assumption on the target.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Ziyue Cai, Weiye Gan, Borui Miao, Zuoqiang Shi. 2026-09-12. Accelerating Langevin Dynamics with Weighted Laplacian Flow: Global Well-Posedness and Exponential Convergence. https://arxiv.org/abs/2610.35892

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

Quantitative Propagation of Chaos for Singular Interacting Particle Systems Driven by Fractional Brownian Motion

We consider interacting systems particle driven by i.i.d. fractional Brownian motions, subject to irregular, possibly distributional, pairwise interactions. We show propagation of chaos and mean field convergence to the law of the associated McKean--Vlasov equation, as the number of particles $N\to\infty$, with quantitative sharp rates of order $N^{-1/2}$. Our results hold for a wide class of possibly time-dependent interactions, which are only assumed to satisfy a Besov-type regularity, related to the Hurst parameter $H\in (0,+\infty)\setminus \mathbb{N}$ of the driving noises. In particular, as $H$ decreases to $0$, interaction kernels of arbitrary singularity can be considered, a phenomenon frequently observed in regularization by noise results. Our proofs rely on a combinations of Sznitman's direct comparison argument with stochastic sewing techniques.

math.PR↗

Power rate of convergence of discrete curves: framework and applications

We develop a framework for power-law rates of convergence of random discrete curves to Schramm Loewner Evolution (SLE). We show that a power-law rate of convergence of a suitable martingale observable gives a power-law rate of convergence of the interface to SLE, provided the discrete curves satisfy the Kemppainen-Smirnov crossing condition; the proof uses known estimates on the derivative of the SLE map. We apply the framework to the exploration process of critical site percolation on the triangular lattice, obtaining a power-law rate of convergence to SLE$_6$.

math.PR↗

Refining Concentration Bounds for Gaussian Quadratic Chaos with Applications

This paper studies concentration of measure for Gaussian quadratic chaos in the non-asymptotic regime. We begin by tightening the classic Hanson-Wright inequality, where we increase its absolute constant $κ$ from 0.125 to at least 0.1457 in the symmetric case. We present a sharper version of Laurent-Massart inequality (LMI), which increases $κ$ from the largest available value of $1-\frac{\sqrt{3}}{2}\approx 0.134$ to $\frac{9-\sqrt{17}}{32}\approx 0.152$ in the positive-semidefinite case. We develop a sequence of inequalities indexed by $m\ge1$ that involve Schatten norms of the underlying matrix. The case $m=1$ recovers HWI. The bounds undergo a phase transition in the sense that if the tail parameter is smaller than a critical threshold $τ_c$, then $m=1$ is the tightest, whereas if it is larger than $τ_c$, then $m=\infty$ is the tightest. This leads to a novel bound called the $m_\infty$ bound. All HWI, LMI and the $m_\infty$ inequality rely on Markov's inequality. By avoiding Markov's inequality, we introduce the strong and weak $χ^2$ inequalities. We provide a list of concentration bounds that depend only on the operator norm in the positive-definite case, namely, the $m_\infty$ bound, the weak $χ^2$ bound, relaxed versions of HWI and LMI, and the so-called large-deviations bound. The sharpest among these is shown to be always either the $m_\infty$ bound or the weak $χ^2$ bound. By utilizing conditioning, modified versions of $m_\infty$, $χ^2$, and HW inequalities of various orders are proposed, all of which can be significantly tighter than the original inequalities. By leveraging the Ruben-Kotz expansion for the distribution of a positive-definite quadratic form, we introduce the Ruben-Kotz inequalities (RKIs) for general symmetric quadratic forms. A key advantage of RKIs is that they provide an explicit upper bound on the approximation gap.

math.PR↗