Local Vertex Colouring Graph Neural Networks
Authors: Shouheng Li, Dongwoo Kim, Qing Wang
ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate SGN on two prediction tasks: vertex classification and graph classification. SGN is implemented using Pytorch and Pytorch Geometric (Fey & Lenssen, 2019). Experiments are run on a single NVIDIA RTX A6000 GPU with 48GB memory. |
| Researcher Affiliation | Collaboration | 1School of Computing, Australian National University, Canberra, Australia 2Data61, CSIRO, Canberra, Australia 3CSE & GSAI, POSTECH, Pohang, South Korea. |
| Pseudocode | No | The paper presents mathematical update rules in Equations 7 and 8, but not in a formal pseudocode block or algorithm structure. |
| Open Source Code | Yes | Our code is available at https://github.com/seanli3/lvc. |
| Open Datasets | Yes | We use three citation graphs, CORA, CITESEER and PUBMED, and two Amazon co-purchase graphs, COMPUTERS and PHOTO. ... We also use two Wikipedia graphs, CHAMELEON and SQUIRREL, and two webpage graphs, TEXAS and CORNELL, from Web KB (Pei et al., 2020). ... using four molecular datasets: D&D (Dobson & Doig, 2003), PROTEINS (Borgwardt et al., 2005), NCI1 (Wale et al., 2008) and ENZYMES (Schomburg et al., 2004). We also include a social dataset IMDB-BINARY. |
| Dataset Splits | Yes | We adopt the same experimental setup as He et al. (2021), where each dataset is randomly split into train/validation/test set with the ratio of 60%/20%/20%. |
| Hardware Specification | Yes | Experiments are run on a single NVIDIA RTX A6000 GPU with 48GB memory. |
| Software Dependencies | No | SGN is implemented using Pytorch and Pytorch Geometric (Fey & Lenssen, 2019). No specific version numbers are provided for these software dependencies. |
| Experiment Setup | Yes | We perform a hyperparameter search on four parameters in the following ranges: number of layers {1, 2, 3, 4, 5}, dropout probability {0.2, 0.5, 0.7, 0.9}, δ {1, 2, 3, 4}, and hidden layer dimension {64, 128}. |