CausalNET

This repository provides the source code and appendix for CausalNET (IJCAI'24).

Environments

pip install -r requirements.txt

Experiments

Execute the following steps to replicate our results on the two real datasets (i.e., Micro-24 and Micro-25):

cd ./model
conda activate python_envs_name
## before running the following commands, please replace the directory path in '...' by your own settings
## '...': Main.py: line22 & line24
python -u main.py -g 1 -task test_M24 -opt ./configs/config_m24.yaml
python -u main.py -g 2 -task test_M25 -opt ./configs/config_m25.yaml

Note: 
(1) the DAG (causal graph) files will be saved in the subdirectory named './dags/final_prob/'.
(2) the DAG file for 'dataset_name' will be named as 'dataset_name_i.npy'.

Based on the hyper-parameter settings we provided, the estimated training duration for CausalNET is expected to be 2～6 hours (depending on the status of the hardware devices).

Acknowledgement

Thanks to these excellent open source projects:

• TrustworthyAI
• Topological Hawkes Process (TNNLS'22)
• Transformer Hawkes Process (ICML'20)
• CUTS: NEURAL CAUSAL DISCOVERY FROM IRREGULAR TIME-SERIES DATA (ICLR'23)

Citation

If you find the repository helpful, please cite the following paper:

@inproceedings{hua2024causalnet,
  title={CausalNET: Unveiling Causal Structures on Event Sequences by Topology-Informed Causal Attention},
  author={Hua, Zhu and Hong, Huang and Kehan, Yin and Zejun, Fan and Hai, Jin and Bang, Liu},
  booktitle={Proceedings of the 33rd International Joint Conference on Artificial Intelligence},
  year={2024}
}

Contact

Please feel free to contact us if you have questions, or need explanations: huazhu@hust.edu.cn.