Fairness in Network Representation by Latent Structural Heterogeneity in Observational Data

Authors: Xin Du, Yulong Pei, Wouter Duivesteijn, Mykola Pechenizkiy3809-3816

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct randomly controlled experiments on synthetic datasets and verify our methods on real-world datasets. Both quantitative and quantitative results show that our method is effective to recover the fairness of network representations. We design a series of randomized experiments on synthetic and real-world datasets to evaluate our method qualitatively and quantitatively.
Researcher Affiliation Academia Xin Du, Yulong Pei, Wouter Duivesteijn, Mykola Pechenizkiy Technische Universiteit Eindhoven the Netherlands {x.du, y.pei.1, w.duivesteijn, m.pechenizkiy}@tue.nl
Pseudocode No The paper describes algorithms (e.g., beam search) but does not provide pseudocode or a clearly labeled algorithm block within its content. It mentions the beam search algorithm is 'built based on (Duivesteijn, Feelders, and Knobbe 2016, Algorithm 1)', implying it's defined elsewhere.
Open Source Code No The paper does not provide any explicit statement or link to open-source code for the methodology described.
Open Datasets Yes As synthetic datasets, we employ modified versions of the two datasets from (Girvan and Newman 2002). The two datasets are called Karate and Football. We collect the original edge connections including New York Taxi (http://www.nyc.gov/html/tlc/) (K=33, N=1, 013, 845, |V |=265) and Sharing Bike (https://datasf. org/opendata/) (K=27, N=983, 000, |V |=70), as well as the contextual information, e.g. weather records (https://www. ncdc.noaa.gov/) and taxi information.
Dataset Splits No The paper mentions using Node2vec to generate training labels and comparing node representations, but it does not specify explicit training/validation/test dataset splits with percentages, counts, or references to predefined splits for reproducibility of the data partitioning.
Hardware Specification Yes All the experiments are conducted on Linux computing clusters with CPU: 2x Intel Xeon @ 2.1GHz and RAM: 1024GB.
Software Dependencies No The paper mentions building 'the algorithm based on Node2vec (Grover and Leskovec 2016)' but does not provide specific version numbers for Node2vec or any other software components (e.g., programming languages, libraries, frameworks) required for reproducibility.
Experiment Setup Yes The beam search algorithm is built based on (Duivesteijn, Feelders, and Knobbe 2016, Algorithm 1). We set the beam width to 5 and depth to 2.