Reproduce “From GNNs to Trees: Multi-Granular Interpretability for Graph Neural Networks” 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
Jie Yang, Yuwen Wang, Kaixuan Chen, Tongya Zheng, Yihe Zhou, Zhenbang Xiao, Ji Cao, Mingli Song and 1 more
A tree-based framework transforms Graph Neural Networks (GNNs) into interpretable hierarchical structures that capture relationships at multiple granularities, demonstrating enhanced explanatory capabilities while maintaining competitive performance on graph classification tasks through adaptive routing and learnable graph perturbation mechanisms.
Key insights
- Multi-granular interpretability better captures real-world graph relationships that exist at different scales
- Hierarchical compression with perturbation enables balanced exploration of different decision paths
- Adaptive routing helps identify the most relevant features at each granularity level
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.