Research map / Graph learning / GNN architectures
EvenNet: Ignoring Odd-Hop Neighbors Improves Robustness of Graph Neural Networks
Runlin Lei, Zhen WANG, Yaliang Li, Bolin Ding, Zhewei Wei
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 |
|---|---|---|---|
| discard odd-order terms from polynomial filter (GPRGNN -> EvenNet) | GPRGNN · cSBM phi_train=0.75, phi_test=-0.75 | accuracy (%): 53.52 -> 94.59 | Improved |
| discard odd-order terms (GPRGNN -> EvenNet) | GPRGNN · cSBM phi_train=phi_test=0.75 | accuracy (%): 95.93 -> 95.29 | No clear effect |
| discard odd-order terms (GPRGNN -> EvenNet) | GPRGNN · Meta-cora poison attack | accuracy (%): 76.27 -> 77.74 | Improved |
| discard odd-order terms (GPRGNN -> EvenNet) | GPRGNN · MM-citeseer poison attack | accuracy (%): 72.81 -> 73.51 | Improved |
| discard odd-order terms (GPRGNN -> EvenNet) | GPRGNN · clean Cora | accuracy (%): 88.34 -> 87.25 | Hurt |
| discard odd-order terms (GPRGNN -> EvenNet) | GPRGNN · clean Squirrel | accuracy (%): 51.93 -> 49.80 | Hurt |
| discard odd-order terms (GPRGNN -> EvenNet) | GPRGNN · clean Texas | accuracy (%): 93.44 -> 93.77 | Improved |
| hard-remove odd orders (EvenNet) vs soft L1 penalty on odd coefficients (EvenReg) (ablation) | EvenNet · cSBM phi_train=0.50, phi_test=-0.50 | accuracy (%): 78.06 -> 82.57 | Improved |
| add L1 penalty on odd coefficients to full-order filter (GPRGNN -> EvenReg) (ablation) | GPRGNN · cSBM phi_train=0.50, phi_test=-0.50 | accuracy (%): 56.16 -> 78.06 | Improved |
| use even-hop (2-hop) vs 1-hop homophily (ablation) | graph attacks · Meta/MinMax poisoned Cora, Citeseer, ACM | homophily gap: 0.28 -> 0.13 (Meta-Cora) | Improved |
Later papers that compared with this method
4 later accepted papers report EvenNet as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- DeloopSGNN: Revisiting Spectral GNNs Through the Lens of Spatial Aggregation (AAAI 2026 · better on 1 of 2 shared results)
- PC-Conv: Unifying Homophily and Heterophily with Two-Fold Filtering (AAAI 2024 · better on 1 of 1 shared result)
- How Universal Polynomial Bases Enhance Spectral Graph Neural Networks: Heterophily, Over-smoothing, and Over-squashing (ICML 2024 · better on 1 of 1 shared result)
- Unifying Homophily and Heterophily for Spectral Graph Neural Networks via Triple Filter Ensembles (NeurIPS 2024 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Graph Neural Networks Inspired by Classical Iterative Algorithms (ICML 2021 · similarity 0.90)
- Making Classic GNNs Strong Baselines Across Varying Homophily: A Smoothness–Generalization Perspective (NeurIPS 2025 · similarity 0.90)
- Graph Neural Networks Are More Than Filters: Revisiting and Benchmarking from A Spectral Perspective (ICLR 2025 · similarity 0.89)
- Bounding the Expected Robustness of Graph Neural Networks Subject to Node Feature Attacks (ICLR 2024 · similarity 0.89)
- Refining Latent Homophilic Structures over Heterophilic Graphs for Robust Graph Convolution Networks (AAAI 2024 · similarity 0.88)
- What Is Missing For Graph Homophily? Disentangling Graph Homophily For Graph Neural Networks (NeurIPS 2024 · similarity 0.88)
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.
