Skip to content

Latest commit

 

History

History
52 lines (37 loc) · 2.17 KB

README.md

File metadata and controls

52 lines (37 loc) · 2.17 KB

mew-supplementals

This repository contains supplemental materials for the paper of Most-Expected-Winner.

Please refer to our technical report for the additional proofs, experiments, and the solver for RSM voting profiles.

Steps to reproduce all experiment results

1. Hardware setup

We ran all experiments on a Linux machine with the following specs:

  • CPU: 2 x Intel Xeon Platinum 8268 24-core 2.9GHz Processor
  • Memory: 384GB RAM

2. Experiment environment setup

The experiment code was in Python and all experiments were conducted on Linux.

Below are the steps to prepare for an independent Python environment with required packages:

  1. Install Miniconda
  2. conda env create -f environment.yml

3. Run all experiments

Execute run_experiments.sh

4. Generate all figures in paper

Each experiment has its own folder in ./experiments/real/ or ./experiments/synthetic/.

  • Fig. 3(a): experiments/synthetic/posets_for_candidate_pruning/plot.ipynb
  • Fig. 3(b): experiments/synthetic/posets_for_voter_grouping/plot.ipynb
  • Fig. 4: experiments/synthetic/posets_for_parallelization/plot.ipynb
  • Fig. 5: experiments/synthetic/posets/plot.ipynb
  • Fig. 6: experiments/synthetic/posets_for_cover_width/plot.ipynb
  • Fig. 7: experiments/synthetic/ppwm/plot.ipynb
  • Fig. 8: experiments/synthetic/pp/plot.ipynb
  • Fig. 9: experiments/synthetic/pc/plot.ipynb
  • Fig. 10: experiments/synthetic/tr/plot.ipynb
  • Fig. 11: experiments/synthetic/mallows/plot.ipynb
  • Fig. 12: experiments/synthetic/rsm/plot.ipynb
  • Fig. 13(a): experiments/synthetic/mallows_pp/plot.ipynb
  • Fig. 13(b): experiments/synthetic/mallows_tr/plot.ipynb
  • Fig. 13(c): experiments/synthetic/mallows_poset/plot.ipynb
  • Fig. 14: experiments/real/crowdrank/plot.ipynb
  • Fig. 15: experiments/synthetic/mpw_parallel_pluarlity/plot.ipynb
  • Fig. 16: experiments/synthetic/mpw_parallel_borda/plot_new.ipynb

The experiment results in Table 6:

  • MovieLens: experiments/real/movielens/movielens_output.tsv
  • Travel: experiments/real/travel/travel_output.tsv