DIPT: Deep Identification of Propagation Trees in Graph Diffusion

Abstract

Understanding how information or influence propagates through a network—such as during an epidemic outbreak or the spread of misinformation—is a fundamental yet challenging problem. Existing methods have largely focused on source localization, overlooking the reconstruction of propagation trees, i.e., the underlying "who-infected-whom" paths that are essential for interpreting diffusion dynamics. We introduce DIPT (Deep Identification of Propagation Trees), a probabilistic framework that infers propagation trees from observed node diffusion states. DIPT models local influence strengths between nodes and leverages an alternating optimization strategy to jointly learn the diffusion mechanism and reconstruct the propagation structure. Empirical results across five real-world datasets demonstrate that DIPT consistently outperforms existing approaches in accurately reconstructing propagation trees.

BibTeX

@article{memon2025deep,
  title={Deep Identification of Propagation Trees},
  author={Memon, Zeeshan and Ling, Chen and Kong, Ruochen and Seshagiri, Vishwanath and Zufle, Andreas and Zhao, Liang},
  journal={arXiv preprint arXiv:2503.00646},
  year={2025}
}