README.md

Accelerating Approximate Thompson Sampling with Underdamped Langevin Monte Carlo

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:

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:

Contact

Haoyang Zheng, School of Mechanical Engineering, Purdue University

Email: zheng528 at purdue dot edu

More Aboue Me: link