Score: 1

Poisson Midpoint Method for Log Concave Sampling: Beyond the Strong Error Lower Bounds

Published: June 9, 2025 | arXiv ID: 2506.07614v3

By: Rishikesh Srinivasan, Dheeraj Nagaraj

BigTech Affiliations: Google

Potential Business Impact:

Speeds up computer learning by 3x.

Business Areas:
A/B Testing Data and Analytics

We study the problem of sampling from strongly log-concave distributions over $\mathbb{R}^d$ using the Poisson midpoint discretization (a variant of the randomized midpoint method) for overdamped/underdamped Langevin dynamics. We prove its convergence in the 2-Wasserstein distance ($W_2$), achieving a cubic speedup in dependence on the target accuracy ($\epsilon$) over the Euler-Maruyama discretization, surpassing existing bounds for randomized midpoint methods. Notably, in the case of underdamped Langevin dynamics, we demonstrate the complexity of $W_2$ convergence is much smaller than the complexity lower bounds for convergence in $L^2$ strong error established in the literature.

Country of Origin
🇺🇸 United States

Page Count
22 pages

Category
Mathematics:
Probability