ChoiceRank: Identifying Preferences from Node Traffic in Networks

Authors: Lucas Maystre, Matthias Grossglauser

ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate the network choice model on three datasets that are representative of two distinct application domains.
Researcher Affiliation Academia 1School of Computer and Communication Sciences, EPFL, Lausanne, Switzerland.
Pseudocode Yes Algorithm 1 Choice Rank
Open Source Code Yes Our code is publicly available online4. 4See: http://lucas.maystre.ch/choicerank.
Open Datasets Yes Recently, it also released clickstream data from the English version of Wikipedia (Wulczyn & Taraborelli, 2016), providing us with essential ground-truth transition-level data. [...] We also consider a second clickstream dataset from a Hungarian online news portal2. The data is publicly available at http://fimi.ua.ac.be/data/. [...] Next, we consider trip data from Citi Bike, New York City s bicycle-sharing system3. The data is available at https://www.citibikenyc.com/system-data.
Dataset Splits No The paper describes using ground-truth transition data for evaluation and aggregating marginal traffic for model fitting, but it does not specify any explicit train/validation/test dataset splits with percentages, counts, or predefined citations.
Hardware Specification Yes We run 20 iterations of Choice Rank on a dual Intel Xeon E5-2680 v3 machine, with 256 GB of RAM and 6 HDDs configured in RAID 0.
Software Dependencies No The paper mentions implementation in "Rust programming language" but does not provide specific version numbers for Rust or any other ancillary software components, libraries, or solvers used in the experiments.
Experiment Setup Yes We set α = 2.0 and β = 1.0 (these small values simply guarantee the convergence of the algorithm) and declare convergence when λ(t) λ(t 1) 1/n < 10 8.