Research map / Graph learning / Graph algorithms and clustering
Convergent Graph Solvers
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 more heads in the learned linear maps (ablation) | CGS · GVI | MAPE: CGS(16): 1.65; CGS(32): 2.16; CGS(64): 2.36 | Helped |
| Make the transition map input-dependent, including both A and B (ablation) | CGS · GVI | MAPE | Helped |
| Use more GN layers in the parameter-generating network (ablation) | CGS(16) · GVI | MAPE: 7.33 ± 4.59 -> 4.42 ± 3.07 | Helped |
| Use multiple linear iterative maps rather than a single map (ablation) | CGS · Pore network diffusion | test MSE | Helped |
| Adapt the transition map to the input graph rather than use a fixed transition map (ablation) | CGS · GVI | MAPE | Helped |
| Use CGS rather than the strongest listed baseline | CGS · IMDB-B | accuracy: 76.0 ± 3.7 vs. 72.8 ± 2.5 | Hurt |
| Use CGS rather than the strongest listed baseline | CGS · IMDB-M | accuracy: 52.3 ± 2.8 vs. 51.1 ± 2.2 | Hurt |
| Use CGS rather than the strongest listed baseline | CGS · MUTAG | accuracy: 91.6 ± 6.5 vs. 89.4 ± 5.6 | No clear effect |
| Use CGS rather than the strongest listed baseline | CGS · PROTEINS | accuracy: 77.4 ± 4.9 vs. 76.3 ± 6.3 | No clear effect |
| Use CGS rather than the strongest listed baseline | CGS · PTC | accuracy: 64.9 ± 7.5 vs. 64.7 ± 6.4 | No clear effect |
| Use CGS rather than the strongest listed baseline | CGS · NCI1 | accuracy: 82.7 ± 1.7 vs. 77.6 ± 2.0 | Hurt |
Closest papers by meaning
- Learning Steady-States of Iterative Algorithms over Graphs (ICML 2018 · similarity 0.89)
- A Convergence Analysis of Gradient Descent on Graph Neural Networks (NeurIPS 2021 · similarity 0.88)
- Are Graph Neural Networks Optimal Approximation Algorithms? (NeurIPS 2024 · similarity 0.87)
- How Graph Neural Networks Learn: Lessons from Training Dynamics (ICML 2024 · similarity 0.86)
- Almost Surely Asymptotically Constant Graph Neural Networks (NeurIPS 2024 · similarity 0.86)
- Two Heads Are Better Than One: Boosting Graph Sparse Training via Semantic and Topological Awareness (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.