Graph Neural Networks with a Distribution of Parametrized Graphs
Authors: See Hian Lee, Feng Ji, Kelin Xia, Wee Peng Tay
ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Numerical experiments demonstrate improvements in performance against baseline models on node classification for both heterogeneous and homogeneous graphs. |
| Researcher Affiliation | Academia | 1School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 2School of Physical and Mathematical Sciences, Nanyang Technological University, Singapore. |
| Pseudocode | Yes | Algorithm 1 EMGNN |
| Open Source Code | Yes | The code is available at https://github.com/amblee0306/EMGNN.git. |
| Open Datasets | Yes | All datasets used in this paper are publicly available and open-source. ... The datasets are publicly available at https://github.com/ seongjunyun/Graph_Transformer_Networks. |
| Dataset Splits | Yes | For our evaluation, we specifically selected the datasets Free Solv, ESOL, and Lipophilicity, all of which are designed for graph regression tasks. ... Random split Scaffold split Datasets Free Solv ESOL Lipophilicity Free Solv ESOL Lipophilicity |
| Hardware Specification | Yes | The models were trained on a server equipped with four NVIDIA RTX A5000 GPUs for hardware acceleration. |
| Software Dependencies | No | The paper mentions specific software libraries like "Cog DL library", "DGL library", and "pygcn" for baselines, but does not provide specific version numbers for these software dependencies or for their own implementation. |
| Experiment Setup | Yes | The hidden units are set to 64 for all models. The hyperparameters of Se HGNN and Simple-HGN are as in the respective repository. ... the number of MCMC iterations Nmc is set to be 15000, the T denoting the number of EM iterations and T is tuned by searching on the following search spaces: [10, 15, 20, 25, 30]. |