Augmentations in Hypergraph Contrastive Learning: Fabricated and Generative

Authors: Tianxin Wei, Yuning You, Tianlong Chen, Yang Shen, Jingrui He, Zhangyang Wang

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

Reproducibility Variable Result LLM Response
Research Type Experimental The experimental findings include: (i) Among fabricated augmentations in Hyper GCL, augmenting hyperedges provides the most numerical gains, implying that higher-order information in structures is usually more downstream-relevant; (ii) Generative augmentations do better in preserving higher-order information to further benefit generalizability; (iii) Hyper GCL also boosts robustness and fairness in hypergraph representation learning. The aforementioned empirical evidences (for generalizability) are drawn from comprehensive experiments on 13 datasets.
Researcher Affiliation Academia University of Illinois Urbana-Champaign, Texas A&M University, University of Texas at Austin
Pseudocode No The paper describes the proposed frameworks and methods in text and uses diagrams (e.g., Figure 1, Figure 3) for illustration, but does not include any structured pseudocode or algorithm blocks.
Open Source Code Yes Codes are released at https://github.com/weitianxin/Hyper GCL.
Open Datasets Yes We examine our methods on the most comprehensive hypergraph benchmarks with 13 data sets, with statistics shown in Table 1. Please refer to Appendix C for detailed information. We use several widely adopted benchmark datasets, including Cora, Citeseer, Pubmed, Cora-CA, DBLP-CA (from [51]); Zoo, 20News, Mushroom, NTU2012, ModelNet40, Yelp, House, Walmart (from [3]) which are publicly available. So we introduce three newly curated hypergraph data sets: German [48], Recidivism [49] and Credit [50].
Dataset Splits Yes By default, we split the data into training/validation/test samples using (10%/10%/80%) splitting percentages.
Hardware Specification No The provided paper text does not specify explicit hardware details such as GPU models, CPU types, or memory specifications used for running the experiments. While the checklist mentions Appendix C contains this information, the content of Appendix C is not provided in the given text.
Software Dependencies No The provided paper text does not specify software dependencies with version numbers (e.g., Python version, specific library versions). While the checklist mentions Appendix C has implementation details, the content of Appendix C is not included in the provided text.
Experiment Setup No The paper states: "All the implementation details are listed in Appendix C. More experiments of the hyperparameters study are given in Appendix A." However, the content of Appendix A and C is not provided in the main paper text, so specific experimental setup details are not directly accessible in the provided text.