Reproduce “Source Localization of Graph Diffusion via Variational Autoencoders for Graph Inverse Problems” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Chen Ling, Junji Jiang, Junxiang Wang, Liang Zhao
This paper introduces the Source Localization Variational Autoencoder (SL-VAE), a probabilistic framework that uses a deep generative model to identify the origin of diffusion in networks from observed patterns. SL-VAE consistently outperforms existing methods on real-world datasets and demonstrates linear scalability with graph size, providing a flexible approach that quantifies uncertainty in source localization.
Key insights
- Source localization, an inverse problem, is inherently probabilistic, and a generative model can better capture the uncertainty and multiple potential source configurations.
- A VAE framework, incorporating an encoder, decoder, and a forward diffusion model, can learn the complex distribution of possible sources given an observed diffusion pattern.
- Combining a standard VAE objective with a forward diffusion consistency loss ensures the reconstructed sources align with the observed diffusion, even when the underlying diffusion model is unknown or complex.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.