How to Sample Approval Elections?

Authors: Stanisław Szufa, Piotr Faliszewski, Łukasz Janeczko, Martin Lackner, Arkadii Slinko, Krzysztof Sornat, Nimrod Talmon

IJCAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We use six datasets. Five of them are generated using our statistical cultures and consist of 100 candidates and 1000 voters (except for the experiments related to the cohesiveness level, where we have 50 candidates and 100 voters, due to computation time). Our visualizations are shown in Figures 3 and 4.
Researcher Affiliation Academia 1AGH University, Poland 2DBAI, TU Wien, Austria 3University of Auckland, New Zealand 4Ben-Gurion University of the Negev, Israel
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code No The paper mentions a third-party library 'abcvoting' and the 'Pabulib' dataset, but does not provide specific access to the source code for the methodology described in this paper.
Open Datasets Yes The sixth dataset uses real-life participatory budgeting data and contains 44 elections from Pabulib [Stolicki et al., 2020], where for each (large enough) election we randomly selected a subset of 50 candidates and 1000 voters.
Dataset Splits No The paper does not provide specific details on training, validation, and test splits for the datasets used in the experiments. It only mentions the overall number of candidates and voters for generated elections and subsets for Pabulib.
Hardware Specification Yes Less than 1 second on a single core (Intel Xeon Platinum 8280 CPU @ 2.70GH) of a 224 core machine with 6TB RAM.
Software Dependencies No The paper mentions using the 'abcvoting library' [Lackner et al., 2021] and the 'Gurobi ILP solver', but does not provide specific version numbers for these software components. The link for abcvoting is to a 'Current version' and not a fixed one.
Experiment Setup Yes Concretely, we consider the following four statistics: Max. Approval Score. Cohesiveness Level. Voters in Cohesive Groups. PAV Runtime. We use six datasets. Five of them are generated using our statistical cultures and consist of 100 candidates and 1000 voters (except for the experiments related to the cohesiveness level, where we have 50 candidates and 100 voters, due to computation time). The Euclidean Model uses radius in (0.0025, 0.25) and (0.005, 0.5).