SpaceGNN: Multi-Space Graph Neural Network for Node Anomaly Detection with Extremely Limited Labels
February 5, 2025 · View on GitHub
Tested Environment
- Ubuntu 22.04
- Python 3.7.16
- Sklearn 1.0.2
- Pytorch 1.12.1
- Numpy 1.21.6
- Torch_geometric 2.3.1
- Scipy 1.7.3
- Dgl 0.9.0
Datasets
Download data files from GADbench (or import from dgl) and put them in datasets/.
Directory Structure
├── datasets
│ ├── weibo
│ │ ├── weibo (different datasets may have different types)
│ ├── main.py
│ ├── utils.py
│ ├── name.py
Use main.py to generate training/validation/test set.
Example
python main.py --data weibo
Experiments
Parameters As described in paper.
Example
python main.py --data weibo