Bootstrapping Heterogeneous Graph Representation Learning via Large Language Models: A Generalized Approach
Authors: Hang Gao, Chenhao Zhang, Fengge Wu, Changwen Zheng, Junsuo Zhao, Huaping Liu
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Theoretical analysis and experimental validation have demonstrated the effectiveness of our method. Experiments Comparison with State of the Art methods. Datasets. We utilized existing commonly used heterogeneous and homogeneous graph representation learning datasets, as well as more challenging heterogeneous graph datasets that we newly constructed. Table 2: Comparative experiment results for IMDB and DBLP datasets. |
| Researcher Affiliation | Academia | 1National Key Laboratory of Space Integrated Information System, Institute of Software, Chinese Academy of Sciences. 2University of Chinese Academy of Sciences. 3Tsinghua University. Corresponding author, EMAIL. |
| Pseudocode | No | No explicit pseudocode or algorithm blocks were found in the main body of the paper. The methodology is described using prose and mathematical equations. |
| Open Source Code | Yes | Code, Datasets and Appendix https://github.com/zch65458525/GHGRL/tree/main |
| Open Datasets | Yes | Code, Datasets and Appendix https://github.com/zch65458525/GHGRL/tree/main. Specifically, we employed the IMDB, DBLP, ACM (Zhang et al. 2019) and Wiki-CS (Mernyei and Cangea 2020) datasets. |
| Dataset Splits | Yes | Datasets IMDB (10% Training) IMDB (40% Training) DBLP (10% Training) DBLP (40% Training). Additionally, we adjusted the proportion of training data in the datasets to compare test results under different conditions. |
| Hardware Specification | No | No specific hardware details (like GPU models, CPU types, or cloud instance specifications) are provided in the main text of the paper. The paper states: "The specific experimental setup, including hyperparameters and the environment used, is detailed in Appendix D.", but Appendix D is not provided. |
| Software Dependencies | No | The paper mentions using 'Llama 3 (Dubey et al. 2024)' as the backbone LLM, but does not provide specific version numbers for software libraries or dependencies used in the implementation. It refers to the LLM model itself rather than specific software versions. |
| Experiment Setup | No | The specific experimental setup, including hyperparameters and the environment used, is detailed in Appendix D. The parameters mfmt and mcont are hyperparameters controlling the size of Φfmt and Φcont, respectively. where α be a hyperparameter to control the proportion of original node features. However, specific values for these hyperparameters are not provided in the main text, and Appendix D is not available. |