Metraipolis › archive › dc_ebfc58e2c22a
Sharp dimensional analysis of midpoint methods for Langevin sampling — arXiv:2610.06308
- Source
- arXiv
- Original
- https://arxiv.org/abs/2610.06308
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-06T05:32:28.959Z
- SHA-256 of the archived text
29c0701e85bb9434e1a2a19b037ea83c7ef499f96a55ca2830eed107db1634c2- Stable URL of this page
https://metraipolis.com/documento/dc_ebfc58e2c22a
How to cite
Sharp dimensional analysis of midpoint methods for Langevin sampling — arXiv:2610.06308 — arXiv, https://arxiv.org/abs/2610.06308 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ebfc58e2c22a (Metraipolis archive dc_ebfc58e2c22a, sha256 29c0701e85bb9434e1a2a19b037ea83c7ef499f96a55ca2830eed107db1634c2)
BibTeX
@misc{dcebfc58e2c22a,
title = {Sharp dimensional analysis of midpoint methods for Langevin sampling — arXiv:2610.06308},
howpublished = {arXiv},
url = {https://arxiv.org/abs/2610.06308},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ebfc58e2c22a; sha256 29c0701e85bb9434e1a2a19b037ea83c7ef499f96a55ca2830eed107db1634c2},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.06308. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
We study deterministic and randomized midpoint discretizations of Langevin dynamics for a target $\pi \propto e^{-V}$, where $0 \prec \alpha I\preceq\nabla^2V\preceq\beta I$ and $\kappa=\beta/\alpha$. To achieve $\sqrt\alpha\,W_2\leqslant\varepsilon$, we show that deterministic Heun uses at most $\widetilde O(\kappa^{4/3}d^{1/3}\varepsilon^{-2/3})$ gradient queries, and underdamped exponential midpoint uses $\widetilde O(\kappa^{5/4}d^{1/4}\varepsilon^{-1/2})$. The proofs exploit cancellation at stationarity and smoothing using techniques from Malliavin calculus, outperforming previous upper bounds based on standard couplings. At bounded condition number, a lower bound matches the $d$ and $\varepsilon$ powers of both deterministic methods. To contrast, for the randomized midpoint methods and Poisson midpoint with at least two grid points (both overdamped and underdamped variants), a simple Gaussian calculation yields a lower bound $d^{1/3}\varepsilon^{-1/3}$ to get an $\varepsilon$-close sample despite starting at a benign initialization. This shows surprisingly that in high dimensions, deterministic discretizations can outperform their random counterparts.
Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Statistics Theory (math.ST) ; Data Structures and Algorithms (cs.DS); Machine Learning (cs.LG); Numerical Analysis (math.NA).
Pagina del lavoro: https://arxiv.org/abs/2610.06308 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-06T05:27:06.032Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).