Hyperbolic Graph Diffusion Model
Authors: Lingfeng Wen, Xuan Tang, Mingjie Ouyang, Xiangxiang Shen, Jian Yang, Daxin Zhu, Mingsong Chen, Xian Wei
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments show that HGDM achieves better performance in generic graph and molecule generation benchmarks, with a 48% improvement in the quality of graph generation with highly hierarchical structures. (Abstract) and Experimental Results In this section, we evaluate our method on generic graph datasets and molecular datasets, then compare it with the baselines. |
| Researcher Affiliation | Academia | 1Mo E Engineering Research Center of Hardware/Software Co-design Technology and Application, East China Normal University 2School of Communication and Electronic Engineering, East China Normal University 3School of Geospatial Information, Information Engineering University, China; 4Quanzhou Normal University |
| Pseudocode | Yes | Algorithm 1: Hyperbolic PC sampling (VE SDE) and Algorithm 2: Hyperbolic PC sampling (VP SDE) |
| Open Source Code | Yes | Code is available at https://github.com/LF-WEN/HGDM |
| Open Datasets | Yes | Generic Graph Generation Experimental Setup We validate HGDM on four generic graph datasets: (1) Ego-small, 200 small ego graphs drawn from larger Citeseer network dataset (Sen et al. 2008), (2) Community-small, 100 randomly generated community graphs (Jo, Lee, and Hwang 2022), (3) Enzymes, 587 protein graphs which represent the protein tertiary structures of the enzymes from the BRENDA database (Schomburg et al. 2004), and (4) Grid, 100 standard 2D grid graphs (Jo, Lee, and Hwang 2022). and Molecule Generation Datasets Following the GDSS (Jo, Lee, and Hwang 2022), we tested our model on the QM9 (Ramakrishnan et al. 2014) and ZINC250k (Irwin et al. 2012) datasets |
| Dataset Splits | No | The paper describes the datasets used and mentions comparing generated graphs to 'test graphs', but does not specify the explicit training, validation, and test splits (e.g., percentages or counts) or the methodology for creating them. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware used to run the experiments, such as GPU models, CPU types, or cloud computing instance specifications. |
| Software Dependencies | No | The paper mentions using the 'RDKit library' and refers to a PyTorch GAT implementation, but it does not provide specific version numbers for these or any other software dependencies needed for replication. |
| Experiment Setup | No | The paper describes general training stages and components (e.g., training HVAE, training two score models, loss functions) but does not provide specific hyperparameter values such as learning rates, batch sizes, or number of epochs, which are crucial for replicating the experimental setup. |