Union Subgraph Neural Networks

Authors: Jiaxing Xu, Aihu Zhang, Qingtian Bian, Vijay Prakash Dwivedi, Yiping Ke

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on 18 benchmarks of both graph-level and node-level tasks demonstrate that Union SNN outperforms state-of-the-art baseline models, with competitive computational efficiency.
Researcher Affiliation Academia Nanyang Technological University, Singapore
Pseudocode No The paper describes the model using mathematical equations but does not include any explicitly labeled pseudocode or algorithm blocks.
Open Source Code Yes Our code is available at https://github.com/Angus Monroe/Union SNN.
Open Datasets Yes For graph classification, we use 10 benchmark datasets. Eight of them were selected from the TUDataset (Kersting et al. 2016), including MUTAG, PROTEINS, ENZYMES, DD, FRANKENSTEIN (denoted as FRANK in our tables), Tox21, NCI1 and NCI109. The other two datasets OGBG-MOLHIV and OGBG-MOLBBBP were selected from Open Graph Benchmark (Hu et al. 2020).
Dataset Splits Yes Time cost (hours) for a single run with 10-fold CV, including training, validation, test (excluding preprocessing). (from Table 12 caption) and Graph classification results (average accuracy standard deviation) over 10-fold-CV. (from Table 3 caption).
Hardware Specification No The paper does not provide specific details on the hardware (e.g., CPU, GPU models, memory) used to run the experiments.
Software Dependencies No The paper does not provide specific version numbers for software dependencies or libraries used for implementation (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup No The implementation details of our experiments are available in our ar Xiv version (Xu et al. 2023). This indicates that comprehensive setup details like hyperparameters are not in the provided document.