/maxgap_bandit

Primary LanguagePythonMIT LicenseMIT

Maxgap Bandit: Adaptive Algorithms for Approximate Ranking

This repository contains the code for algorithms and experiments in the NeurIPS 2019 paper - Maxgap Bandit: Adaptive Algorithms for Approximate Ranking. The paper can be found at https://arxiv.org/abs/1906.00547.

To generate plots for stopping times, run the following commands:

  • cd stopping_times
  • python generate_data.py
  • python stopping_time_plot.py

To generate plots for simulated data, run the following commands:

  • cd simulated_data
  • python generate_data.py
  • python pulls_plot.py
  • python animation_movie.py
  • python mistake_probability_plot.py

To generate plots for Streetview data, run the following commands:

  • cd streetview_data
  • python generate_data.py (takes almost a day to run)
  • python mistake_probability_plot.py