Adaptive Source Localization on Complex Networks via Conditional Diffusion Model
Keywords: Diffusion Model, Knowledge Informed Machine Learning, Source Localization, Complex Network
TL;DR: We propose a diffusion-based source localization method that can directly applied to real-world data in zero-shot manner after pretraining on simulation data with known propagation patterns and simple network topology.
Abstract: Network propagation issues like the spread of misinformation, cyber threats, or infrastructure breakdowns are prevalent and have significant societal impacts. Identifying the source of such propagation by analyzing snapshots of affected networks is crucial for managing crises like disease outbreaks and enhancing network security. Traditional methods rely on metrics derived from network topology and are limited to specific propagation models, while deep learning models face the challenge of data scarcity. We propose \textbf{ASLDiff}~(\textbf{A}daptive \textbf{S}ource \textbf{L}ocalization \textbf{Diff}sion Model), a novel adaptive source localization diffusion model to achieve accurate and robust source localization across different network topologies and propagation modes by fusing the principles of information propagation and restructuring the label propagation process within the conditioning module. Our approach not only adapts to real-world patterns easily without abundant fine-tuning data but can also generalize to different network topologies easily. Evaluations of various datasets demonstrate ASLDiff's superior effectiveness, accuracy, and adaptability in real-world applications, showcasing its robust performance across different localization scenarios. The code can be found at https://anonymous.4open.science/r/ASLDiff-4FE0.
Primary Area: learning on graphs and other geometries & topologies
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 7657
Loading