Skip to content

The GitHub repository for "Accelerating Approximate Thompson Sampling with Underdamped Langevin Monte Carlo", AISTATS 2024.

License

Notifications You must be signed in to change notification settings

haoyangzheng1996/ts_ulmc

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

42 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Accelerating Approximate Thompson Sampling with Underdamped Langevin Monte Carlo

Paper Dependencies License: MIT

Part of experimental code for "TS-ULMC".

@article{zheng2024accelerating,
  title={{Accelerating Approximate Thompson Sampling with Underdamped Langevin Monte Carlo}},
  author={Zheng, Haoyang and Deng, Wei and Moya, Christian and Lin, Guang},
  booktitle={International Conference on Artificial Intelligence and Statistics},
  pages={2611--2619},
  year={2024},
  organization={PMLR}
}

Prerequisites

Please refer to "environment.yml"

Usage

For our method, please run:

python3 ts_underdamp.py

For the baseline, please run:

python3 ts_overdamp.py

Results: image

Further example

To see the results in the appendix, we first used google-maps-scraper to collect Google Maps reviews from several restaurants.

The data were further processed and saved in the folder "./data/" as txt files.

Then please run:

python3 restaurant_plot.py --n_round 200 --batch_size 5 --reward_size 10 --step_size 1e-2

Results: image

Contact

Haoyang Zheng, School of Mechanical Engineering, Purdue University

Email: zheng528 at purdue dot edu

More Aboue Me: link

About

The GitHub repository for "Accelerating Approximate Thompson Sampling with Underdamped Langevin Monte Carlo", AISTATS 2024.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages