Skip to content

Top-m arm identification in linear bandit algorithms

License

Notifications You must be signed in to change notification settings

clreda/linear-top-m

Repository files navigation

linear-top-m

Code associated with AISTATS 2021 paper "Top-m identification for linear bandits". The code was successfully run on Python 2.7.14+, on Linux Debian 8. To cite the paper, please use:

@inproceedings{reda2021top,
  title={Top-m identification for linear bandits},
  author={R\'{e}da, Cl\'{e}mence and Kaufmann, \'{E}milie and Delahaye-Duriez, Andr\'{e}e},
  booktitle={Proceedings of the $24^{th}$ International Conference on Artificial Intelligence and Statistics (AISTATS) 2021},
  volume={130},
  year={2021}
}

TL;DR

  • Install necessary packages:
python -m pip install --force-reinstall pip==9.0.1
pip install -r requirements.txt
//most up-to-date version of https://github.com/regulomics/expansion-network according to GitHub
git clone https://github.com/clreda/expansion-network expansion-network
  • Data for the instance of drug repurposing for epilepsy (concept id/CID: C0014544) is stored in DR_Data/ (data is from LINCS L1000 and RepoDB, and data from GEO accession number GSE77578).

  • In order to reproduce the exact figures shown in the paper, run generate_pdf.py which uses files in folder Results/.

Command options

  • default argument values are written in JSON in file args.json. Necessary arguments are:

    • small_N for data=linear or logistic
    • small_K for data=classic or linear or logistic
    • vr for data=linear or logistic
    • omega for data=classic
    • T_init for bandit=TrueUniform
    • bandit
    • data
    • m
    • beta
  • Adding "--verbose 1" allows to display comments and description of selected arms and obtained rewards.

Modifying the code

Add a new bandit algorithm

  • Check file bandits.py. You need to create a sub-instance of ExploreMBandit and add it to the bandit_factory. Then you will be able to invoke it from the terminal with option "--bandit".

Add a new type of data

  • Check file data.py. You need to modify function create_scores_features. Then you will be able to invoke it from the terminal with option "--data".

Add a new threshold function

  • Check file betas.py. You need to create a function similar to the AlphaDependentBeta, LUCB1Beta, ... and add it to the beta_factory. Then you will be able to invoke it from the terminal with option "--beta".

Add a new problem type

  • Check file problems.py. You need to create a sub-instance of GenericProblem and add it to the problem_factory. Then you will be able to invoke it from the terminal with option "--problem".

About

Top-m arm identification in linear bandit algorithms

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published