Source Code of ICDE'24 submitted paper "BOURNE: Bootstrapped Self-supervised Learning Framework for Unified Graph Anomaly Detection"
- python==3.8.13
- torch_geometric==2.3.0
- matplotlib==3.5.1
- networkx==3.1
- numpy==1.21.6
- scipy==1.8.0
- sklearn==0.24.1
- torch==2.0.0
- tqdm==4.65.0
To train and evaluate on Cora:
cd NAD
python train_node.py --dataset cora --layer_sizes 256 --epochs 500 --batch_size 2000 --lr 0.001 --alpha 1.0 --beta 0.4 --eval_rounds 200 --cudaID 0
To train and evaluate on Cora:
cd EAD
python train_edge.py --dataset cora --layer_sizes 256 --epochs 500 --batch_size 2000 --lr 0.001 --alpha 1.0 --beta 0.4 --eval_rounds 200 --cudaID 0
@inproceedings{liu2024bourne,
title={Bourne: Bootstrapped self-supervised learning framework for unified graph anomaly detection},
author={Liu, Jie and He, Mengting and Shang, Xuequn and Shi, Jieming and Cui, Bin and Yin, Hongzhi},
booktitle={2024 IEEE 40th International Conference on Data Engineering (ICDE)},
pages={2820--2833},
year={2024},
organization={IEEE}
}