Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
When Hypergraph Meets Heterophily: New Benchmark Datasets and Baseline
Authors: Ming Li, Yongchun Gu, Yi Wang, Yujie Fang, Lu Bai, Xiaosheng Zhuang, Pietro Liò
AAAI 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments conducted on synthetic and benchmark datasets highlight the challenges current HNNs face with heterophilic hypergraphs, while showcasing that Hyper UFG performs competitively and often outperforms many existing models in such scenarios. Overall, our study underscores the urgent need for further exploration and development in this emerging field, with the potential to inspire and guide future research in HHL. |
| Researcher Affiliation | Academia | 1Zhejiang Key Laboratory of Intelligent Education Technology and Application, Zhejiang Normal University 2Zhejiang Institute of Optoelectronics 3School of Computer Science and Technology, Zhejiang Normal University 4School of Artificial Intelligence, and Engineering Research Center of Intelligent Technology and Educational Application, Ministry of Education, Beijing Normal University 5Department of Mathematics, City University of Hong Kong 6Department of Computer Science and Technology, Cambridge University EMAIL, EMAIL, EMAIL, EMAIL |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. The methods are described using mathematical formulations and textual descriptions. |
| Open Source Code | Yes | HHL Repository https://kellysylvia77.github.io/HHL Appendix https://mingli-ai.github.io/HHL.pdf |
| Open Datasets | Yes | Amazon-ratings (co-purchasing). This dataset is derived from the Amazon product co-purchasing network metadata2, sourced from SNAP Datasets (Jure 2014). It consists of nodes representing a diverse array of products such as books, music CDs, DVDs, and VHS videotapes... 2https://snap.stanford.edu/data/amazon-meta.html Twitch-gamers (co-create). The Twitch-Gamers network3 is a connected, undirected graph that models relationships among accounts on the Twitch streaming platform... 3http://snap.stanford.edu/data/twitch gamers.html Pokec (co-friendship). Pokec4, the predominant online social networking platform in Slovakia... 4https://snap.stanford.edu/data/soc-Pokec.html |
| Dataset Splits | Yes | For all new benchmarks, we utilize feature vectors, class labels, and ten random splits (40%/20%/40% of nodes per class for training/validation/testing, respectively). |
| Hardware Specification | Yes | The experiments are conducted on a NVIDIA RTX A6000 GPU with 48GB of memory. |
| Software Dependencies | No | The paper mentions the NNI toolkit and provides a link to its documentation but does not provide specific version numbers for any software dependencies, such as programming languages, libraries, or the toolkit itself. |
| Experiment Setup | Yes | We train the model for a total of 1,000 epochs, employing early stopping with a patience threshold of 200 epochs. The baseline results are reproduced using their publicly available code, with hyperparameters set according to the original papers. We utilize grid search to fine-tune the key hyperparameters via the lightweight but powerful toolkit NNI (https://nni.readthedocs.io/en/stable/). Additional details can be found in Appendix A and Appendix B. |