Reproduce “Simplified PCNet with Robustness” 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
Bingheng Li, Xuanting Xie, Haoxiang Lei, Ruiyi Fang, Zhao Kang
Simplified PCNet (SPCNet) is introduced as a Graph Neural Network that performs effectively on both homophilic and heterophilic graphs by employing a cross-receptive filter. The model demonstrates enhanced robustness against structural perturbations and adversarial attacks while maintaining computational efficiency.
Key insights
- A cross-receptive filter, utilizing Poisson-Charlier polynomials, effectively balances local and global information processing to handle both homophilic and heterophilic graph structures.
- Simplifying a complex GNN architecture can lead to improved robustness and efficiency without sacrificing performance.
- Theoretical analysis demonstrates linear stability, indicating that small perturbations in graph structure lead to proportionally small changes in model output, enhancing reliability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.