Skip to content
DraconiticResearch
Search Sources API
201,642 studies31 sourcesUpdated 13 min ago
PreprintCC BY-NC-ND 4.0arXiv

Optimal Scaling of Langevin Proposals with Generalized Acceptance Rules

  • Ritik Soni
  • Dootika Vats
Open at sourcePDFDOICite

Abstract

Langevin-based Markov chain Monte Carlo (MCMC) algorithms use gradient information to improve sampling, particularly in high dimensions. Classical optimal scaling theory for these algorithms has largely focused on the Metropolis-Hastings (MH) acceptance rule. However, there has been a recent surge in acceptance rules beyond MH for applications spanning differential privacy, stochastic MCMC, diffusion models, and molecular dynamics. We develop optimal scaling results for Langevin proposals employed with generalized acceptance rules belonging to a suitable class. For high-dimensional targets, we recover the usual O(d−1/3) scaling, while different acceptance rules lead to different optimal acceptance probabilities.

Related

  1. Scaling limit of the 3D abelian Yang--Mills Langevin dynamics
    Ilya Chevyrev, Yahui Qu, Hao Shen
    arXivPreprintCC0Full textAnalysis & PDEsMathematical physicsarXiv
  2. History-Conditioned Inference of Memory and Stochastic Forcing in Generalized Langevin Equations under Lévy Noise
    Yang Yang, Ting Gao, Xiaoli Chen
    arXivPreprintarXiv licenceFull textStatisticsarXiv
  3. Memory-Conditioned Diffusion Model for Generalized Langevin Dynamics
    Minglei Yang, Sicheng He
    arXivPreprintarXiv licenceFull textMathematical physicsStatisticsarXiv
  4. Dropout Universality: Scaling Laws and Optimal Scheduling at the Edge-of-Chaos
    Lucas Fernandez Sarmiento
    arXivPreprintCC BY 4.0Full textStatisticsCondensed matterarXiv
  5. Near-Optimal Parallel M-Matrix Scaling
    Chih-Hung Liu, Aaron Sidford, Ta-Wei Tu
    arXivPreprintarXiv licenceFull textOptimisation & controlarXiv

Cite

BibTeXRIS
BibTeX
@misc{soni2026optimal,
  title = {{Optimal Scaling of Langevin Proposals with Generalized Acceptance Rules}},
  author = {Ritik Soni and Dootika Vats},
  howpublished = {arXiv},
  year = {2026},
  month = sep,
  doi = {10.48550/arxiv.2609.16941},
  eprint = {2609.16941},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.16941},
  note = {CC BY-NC-ND 4.0},
}
RIS
TY  - UNPB
TI  - Optimal Scaling of Langevin Proposals with Generalized Acceptance Rules
AU  - Ritik Soni
AU  - Dootika Vats
PY  - 2026
DA  - 2026/09/15
PB  - arXiv
DO  - 10.48550/arxiv.2609.16941
UR  - https://arxiv.org/abs/2609.16941
AB  - Langevin-based Markov chain Monte Carlo (MCMC) algorithms use gradient information to improve sampling, particularly in high dimensions. Classical optimal scaling theory for these algorithms has largely focused on the Metropolis-Hastings (MH) acceptance rule. However, there has been a recent surge in acceptance rules beyond MH for applications spanning differential privacy, stochastic MCMC, diffusion models, and molecular dynamics. We develop optimal scaling results for Langevin proposals employed with generalized acceptance rules belonging to a suitable class. For high-dimensional targets, we recover the usual O(d^−1/3) scaling, while different acceptance rules lead to different optimal acceptance probabilities.
ER  -