Hierarchical Graph Capsule Network

Authors: Jinyu Yang, Peilin Zhao, Yu Rong, Chaochao Yan, Chunyuan Li, Hehuan Ma, Junzhou Huang10603-10611

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental studies demonstrate the effectiveness of HGCN and the contribution of each component.
Researcher Affiliation Collaboration 1University of Texas at Arlington 2Tencent AI Lab jzhuang@uta.edu
Pseudocode Yes Algorithm 1: Training process with K latent factors, L capsule layers, and R iterations of routing.
Open Source Code Yes Code: https://github.com/uta-smile/HGCN
Open Datasets Yes Eleven commonly used benchmarks including (i) seven biological graph datasets, i.e., MUTAG, NCI1, PROTEINS, D&D, ENZYMES, PTC, NCI109; and (ii) four social graph datasets, i.e., COLLAB, IMDB-Binary (IMDB-B), IMDB-Multi (IMDB-M), Reddit-BINARY (RE-B), are used in this study.
Dataset Splits Yes perform 10-fold cross-validation for performance evaluation.
Hardware Specification No The paper does not provide specific details about the hardware used for experiments.
Software Dependencies No The paper mentions adopting GCN but does not provide specific version numbers for any software dependencies.
Experiment Setup Yes We set K = 4, R = 3, λ = 0.5, β = 0.1, L = 2, and follow the same settings in previous studies (Ying et al. 2018b) to perform 10-fold cross-validation for performance evaluation.