This repository contains the implementation of algorithms described in the paper "Randomized Confidence Bounds for Partial Monitoring". This branch is a sandbox version, the developer's code is stored on the other branch of the project.
Before you begin, ensure you have met the following requirements:
- Python 3.8
- pip
Follow these steps to set up your environment and run the experiments:
-
Create a Virtual Environment:
python -m venv env source env/bin/activate -
Install Dependencies:
pip install -r requirements.txt
- Cyipopt: If you encounter issues installing cyipopt, ensure you have the latest versions of pip, setuptools, and wheel. You may also need additional system dependencies. For more details, see the Cyipopt Installation Guide.
- Gurobi Alternative: If you prefer not to use Gurobi, you can use PULP as an alternative optimizer. To do this, install PULP using pip install pulp.
- Non-contextual Experiments: To run non-contextual experiments, use the Jupyter notebook
experiment_noncontextual.ipynb. - Contextual Experiments: For contextual experiments, refer to the
experiment_contextual.ipynbnotebook. - Use case Experiments: For the use case, refer to the Use_case folder, approaches C-CBP, C-RandCBP and ExploreFully are in the
utils.pyfile. Specifically, see./use_case/benchmark_use_case2.ipynbscript.
This work was funded through Mitacs with additional support from CIFAR (CCAI Chair). We thank Alliance Canada and Calcul Quebec for access to computational resources and staff expertise consultation. We would like to thank Junpei Komiyama, Taira Tsuchiya, Ian Lienert, Hastagiri P. Vanchinathan and James A. Grant for answering our technical questions and/or providing total/partial access to private code bases of their approaches. We also acknowledge the library pmlib of Tanguy Urvoy that was helpful to implement PM game environments.