Research map / Graph learning / GNN architectures
Graph Neural Dynamics via Learned Energy and Tangential Flows
Moshe Eliasof, Eldad Haber, Carola-Bibiane Schönlieb
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 TANGO energy-descent plus tangential flow on top of GatedGCN | GatedGCN · Graph Property Prediction (Eccentricity) | log10(MSE): 0.6995 -> -1.7419 | Improved |
| Add TANGO to GPS | GPS · Graph Property Prediction (SSSP) | log10(MSE): -3.5990 -> -5.5263 | Improved |
| Add TANGO to GatedGCN | GatedGCN · ZINC-12k | MAE: 0.282 -> 0.128 | Improved |
| Add TANGO to GPS | GPS · ZINC-12k | MAE: 0.070 -> 0.062 | Improved |
| Add TANGO to GatedGCN | GatedGCN · CIFAR10 | accuracy: 67.312 -> 70.894 | Improved |
| Add TANGO to GPS | GPS · CIFAR10 | accuracy: 72.298 -> 75.783 | Improved |
| Add TANGO to GPS | GPS · PATTERN | accuracy: 86.685 -> 87.182 | Improved |
| Add TANGO to GatedGCN | GatedGCN · Peptides-func | AP (%): 58.64 -> 68.92 | Improved |
| Use TANGO_GPS instead of multi-hop DRew-GCN | TANGO_GPS · Peptides-func | AP (%): 71.50 -> 70.21 | Hurt |
| Use TANGO instead of GRIT | TANGO_GPS · CIFAR10 / ZINC-12k | accuracy / MAE: CIFAR10 76.468 vs 75.783; ZINC 0.059 vs 0.062 | Hurt |
| Add TANGO to GPS (runtime cost) | GPS · Questions, 8 layers 256 channels | runtime (ms/epoch): 429.08 -> 694.27 | Hurt |
| Add tangential flow to Dirichlet-energy gradient flow on a barbell graph (ablation) | TANGO · Barbell graph, 50 layers | information propagation | Improved |
Closest papers by meaning
- How Graph Neural Networks Learn: Lessons from Training Dynamics (ICML 2024 · similarity 0.88)
- Towards Dynamic Message Passing on Graphs (NeurIPS 2024 · similarity 0.88)
- When Graph Neural Networks Meet Dynamic Mode Decomposition (ICLR 2025 · similarity 0.88)
- Chasing Label Shifters: A Change-Aware Framework for Dynamic Graph Node Classification (NeurIPS 2026 · similarity 0.87)
- A Dynamical Systems-Inspired Pruning Strategy for Addressing Oversmoothing in Graph Attention Networks (ICML 2025 · similarity 0.87)
- Permutation Equivariant Neural Controlled Differential Equations for Dynamic Graph Representation Learning (NeurIPS 2025 · 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.
