Research map / Graph learning / GNN architectures
From Node Interaction To Hop Interaction: New Effective and Scalable Graph Learning Paradigm
Jie Chen, Zilong Li, Yin Zhu, Junping Zhang, Jian Pu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add nonlinear interaction among precomputed hop features (ablation) | HopGNN · Heterophily datasets (average) | test accuracy: 65.59 -> 69.63 | Improved |
| Add nonlinear interaction among precomputed hop features (ablation) | HopGNN · Homophily datasets (average) | test accuracy: 84.35 -> 84.42 | No clear effect |
| Add hop-order embedding (ablation) | HopGNN · Heterophily datasets (average) | test accuracy: 60.93 -> 69.63 | Improved |
| Add hop-order embedding (ablation) | HopGNN · Homophily datasets (average) | test accuracy: 83.96 -> 84.42 | No clear effect |
| Replace mean fusion with attention fusion (ablation) | HopGNN · Heterophily and homophily datasets (averages) | test accuracy: 69.63 -> 69.81; 84.42 -> 84.60 | No clear effect |
| Replace mean fusion with max fusion (ablation) | HopGNN · Heterophily and homophily datasets (averages) | test accuracy: 69.63 -> 69.50; 84.42 -> 84.39 | No clear effect |
| Replace attention hop interaction with MLP (ablation) | HopGNN · Heterophily datasets (average) | test accuracy: 69.63 -> 67.21 | Hurt |
| Replace attention hop interaction with GCN (ablation) | HopGNN · Heterophily datasets (average) | test accuracy: 69.63 -> 68.18 | Hurt |
| Replace attention hop interaction with SAGE (ablation) | HopGNN · Heterophily and homophily datasets (averages) | test accuracy: 69.63 -> 69.11; 84.42 -> 84.45 | Mixed |
| Add feature-level self-supervised learning | HopGNN · 9 homophily and heterophily datasets (Avg) | test accuracy: 74.56 -> 75.53 | Improved |
| Use supervised contrastive learning | HopGNN · 9 homophily and heterophily datasets (Avg) | test accuracy: 74.56 -> 74.76 | No clear effect |
Closest papers by meaning
- DeltaGNN: Graph Neural Network With Information Flow Control (TPAMI 2026 · similarity 0.90)
- GAMMA: Gated Multi-hop Message Passing for Homophily-Agnostic Node Representation in GNNs (NeurIPS 2025 · similarity 0.89)
- High-Order Pooling for Graph Neural Networks with Tensor Decomposition (NeurIPS 2022 · similarity 0.89)
- K-hop Hypergraph Neural Network: A Comprehensive Aggregation Approach (AAAI 2025 · similarity 0.88)
- GraphTOP: Graph Topology-Oriented Prompting for Graph Neural Networks (NeurIPS 2025 · similarity 0.87)
- Can Classic GNNs Be Strong Baselines for Graph-level Tasks? Simple Architectures Meet Excellence (ICML 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
