Deep Orthogonal Hypersphere Compression for Anomaly Detection

This is the official implementation of Deep Orthogonal Hypersphere Compression for Anomaly Detection, ICLR 2024 (Spotlight).

Dependencies

• python 3.8
• pytorch
• torch-geometric
• torch-sparse
• numpy
• scikit-learn

If you have installed above mentioned packages you can skip this step. Otherwise run:

pip install -r requirements.txt

Reproduce graph data results

To generate results

python demo_DOHSC.py --DS MUTAG --eval True

or

python demo_DO2HSC.py --DS MUTAG --eval True

Reproduce tabular data results

To generate results, please run:

python demo_tabular.py

For running tabular data, the dataset name needs to be revised in corresponding demo files.

Reproduce image data results

To generate results, please run:

python demo_cifar10.py

Reference

If you find this code useful in your research, please consider citing:

@inproceedings{zhang2024deep,
  title={Deep Orthogonal Hypersphere Compression for Anomaly Detection},
  author={Zhang, Yunhe and Sun, Yan and Cai, Jinyu and Fan, Jicong},
  booktitle={The Twelfth International Conference on Learning Representations},
  year={2024}
}