Research map / Graph learning / GNN architectures
EIGNN: Efficient Infinite-Depth Graph Neural Networks
Juncheng Liu, Kenji Kawaguchi, Bryan Hooi, Yiwei Wang, Xiaokui Xiao
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 |
|---|---|---|---|
| Use EIGNN instead of H2GCN | Node classification · Cornell | accuracy (%): 82.22 -> 85.13 | Improved |
| Use EIGNN instead of H2GCN | Node classification · Texas | accuracy (%): 84.76 -> 84.60 | Hurt |
| Use EIGNN instead of H2GCN | Node classification · Wisconsin | accuracy (%): 85.88 -> 86.86 | Improved |
| Use EIGNN instead of H2GCN | Node classification · Chameleon | accuracy (%): 60.30 -> 62.92 | Improved |
| Use EIGNN instead of H2GCN | Node classification · Squirrel | accuracy (%): 40.75 -> 46.37 | Improved |
| Use EIGNN instead of IGNN | Multi-label node classification · PPI | Micro-F1 (%): 97.6 -> 98.0 | Improved |
| Replace IGNN's iterative solver with EIGNN's closed-form, eigendecomposition-based computation | Infinite-depth GNN training · Chains (l=100, n_c=20) | training time per epoch: 27.36ms -> 4.05ms | Improved |
| Replace IGNN's iterative solver with EIGNN's closed-form, eigendecomposition-based computation | Infinite-depth GNN training · Chains (l=200, n_c=20) | training time per epoch: 29.88ms -> 8.03ms | Improved |
| Replace IGNN's iterative solver with EIGNN's closed-form, eigendecomposition-based computation | Infinite-depth GNN training · Chains (l=100, n_c=30) | training time per epoch: 64.94ms -> 6.03ms | Improved |
| Replace IGNN's iterative solver with EIGNN's closed-form, eigendecomposition-based computation | Infinite-depth GNN training · Chains (l=200, n_c=30) | training time per epoch: 60.31ms -> 16.0ms | Improved |
| Use EIGNN instead of IGNN under feature perturbation | Node classification · Cornell (FGSM 0.01) | accuracy (%): 57.30 -> 72.70 | Improved |
| Use EIGNN instead of IGNN under feature perturbation | Node classification · Texas (FGSM 0.01) | accuracy (%): 53.51 -> 74.59 | Improved |
Later papers that compared with this method
4 later accepted papers report EIGNN as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- IGNN-Solver: A Graph Neural Solver for Implicit Graph Neural Networks (TPAMI 2026 · better on 1 of 1 shared result)
- Implicit vs Unfolded Graph Neural Networks (JMLR 2025 · better on 3 of 3 shared results)
- MGNNI: Multiscale Graph Neural Networks with Implicit Layers (NeurIPS 2022 · better on 5 of 6 shared results)
- Optimization-Induced Graph Implicit Nonlinear Diffusion (ICML 2022 · better on 3 of 3 shared results)
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
- Scalable and Effective Implicit Graph Neural Networks on Large Graphs (ICLR 2024 · similarity 0.89)
- Graph Neural Networks Inspired by Classical Iterative Algorithms (ICML 2021 · similarity 0.88)
- Bag of Tricks for Training Deeper Graph Neural Networks: A Comprehensive Benchmark Study (TPAMI 2022 · similarity 0.88)
- Sketch-Augmented Features Improve Learning Long-Range Dependencies in Graph Neural Networks (NeurIPS 2025 · similarity 0.88)
- Graph Distillation with Eigenbasis Matching (ICML 2024 · similarity 0.88)
- On the Generalization of Equivariant Graph Neural Networks (ICML 2024 · similarity 0.86)
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.
