Unsupervised Network Embedding Beyond Homophily

Authors: Zhiqiang Zhong, Guadalupe Gonzalez, Daniele Grattarola, Jun Pang

TMLR 2022 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Comprehensive empirical evaluations on both synthetic and real-world datasets with varying homophily ratios validate the effectiveness of SELENE in homophilous and heterophilous settings showing an up to 12.52% clustering accuracy gain.
Researcher Affiliation Academia Zhiqiang Zhong EMAIL University of Luxembourg Guadalupe Gonzalez EMAIL Imperial College London Daniele Grattarola EMAIL Università della Svizzera italiana Jun Pang EMAIL University of Luxembourg
Pseudocode Yes Algorithm 1: SELf-sup Ervised Network Embedding (SELENE) Framework
Open Source Code Yes Code and data are available at: https://github.com/zhiqiangzhongddu/SELENE
Open Datasets Yes All real-world datasets are available online1. https://pytorch-geometric.readthedocs.io/en/latest/modules/datasets.html
Dataset Splits No The paper mentions evaluating on 'node clustering, node class prediction and link prediction' tasks and for node classification states it 'utilises node labels to train a classifier, after obtaining unsupervised node representations, to predict labels of test nodes'. However, it does not provide specific percentages, sample counts, or explicit methodology for how these splits were performed (e.g., 80/10/10 split, or a random seed for splitting).
Hardware Specification No The paper does not provide specific hardware details such as exact GPU/CPU models, processor types, or memory amounts used for running the experiments.
Software Dependencies No The paper does not provide specific ancillary software details, such as library names with version numbers (e.g., Python 3.8, PyTorch 1.9).
Experiment Setup No The paper mentions hyperparameters like 'probability px, pe' for distortion and 'λ' for the Barlow-Twins loss, but it does not provide their specific values used for the main experimental results. For λ, it references 'default settings as Zbontar et al. (2021)' without stating these values.