Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Code for the paper "Underdamped Langevin MCMC with third order convergence"

by Maximilian Scott, Dáire O'Kane, Andraž Jelinčič and James Foster

The paper is available at arxiv.org/abs/2508.16485.

This repository is based on the Diffrax package by Patrick Kidger, which is available at github.com/patrick-kidger/diffrax. The documentation for Diffrax can be found at docs.kidger.site/diffrax.

For results regarding the strong order of convergence of QUICSORT compared to some other solvers, see notebooks/langevin_order.ipynb.

The main entry point for the Bayesian Logistic Regression example is in mcmc/progressive_run.py. Pregenerated plots can be found in mcmc/progressive_results/good_plots/. For an experiment showing the strong order of convergence on the Bayesian Logistic Regression model, see notebooks/logreg_order.ipynb.

To use the code, clone this repository and install the requirements:

git clone https://github.com/your-username-here/ThirdOrderLMC.git
cd ThirdOrderLMC/
pip install -e .

About

Langevin Monte Carlo experiments for papers "Underdamped Langevin MCMC with third order convergence" and "Single-seed generation of Brownian paths and integrals for adaptive and high order SDE solvers"

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages