Reconstructing graph diffusion history from a single snapshot (KDD 2023)

@inproceedings{qiu2023ditto,
  title={Reconstructing graph diffusion history from a single snapshot},
  author={Ruizhong Qiu and Dingsu Wang and Lei Ying and {H. Vincent} Poor and Yifang Zhang and Hanghang Tong},
  booktitle={Proceedings of the 29th {ACM} {SIGKDD} Conference on Knowledge Discovery and Data Mining},
  year={2023}
}

Dependencies

Our code was tested under the following dependencies:

• CUDA 11.4
• torch==1.7.0
• class-resolver==0.3.10
• torch-scatter==2.0.7
• torch-sparse==0.6.9
• torch-cluster==1.5.9
• torch-geometric==2.0.4
• ndlib==5.1.1
• geopy==2.1.0

Usage

To reproduce our results:

cd scripts
./{method}-{dataset}.sh {device}

• {method}: ditto (ours) / dhrec / cri / gcn / gin.
  • The original code for DHREC is specially for SEIRS, so we provide our implementation of DHREC-PCDSVC for SI & SIR here.
  • The CRI paper did not publish their source code, so we implemented CRI according to their paper.
  • The implementations of GCN and GIN are from PyTorch Geometric.
• {dataset}: ba-si / er-si / oregon2-si / prost-si / farmers-si (BrFarmers) / pol-si / ba-sir / er-sir / oregon2-sir / prost-sir / covid-sir / heb-sir (Hebrew).
  • Notice: As is explained in Section 5.4, {gcn, gin} were evaluated only on {farmers-si, pol-si, covid-sir, heb-sir}.
• {device}: the device for PyTorch.

Other baselines

• BRITS (Cao et al., 2018): paper, code
• GRIN (Cini et al., 2022): paper, code
• SPIN (Marisca et al., 2022): paper, code