Research map / Graph learning / GNN architectures
Reversible and irreversible bracket-based dynamics for deep graph neural networks
Anthony Gruber, Kookjin Lee, Nathaniel Trask
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 |
|---|---|---|---|
| replace black-box NODE with Hamiltonian bracket | NODE · Damped double pendulum | total MAE: 0.0269 -> 0.00369 | Improved |
| use gradient (purely dissipative) bracket instead of Hamiltonian | bracket network · Damped double pendulum | total MAE: 0.00369 -> 0.0208 | Hurt |
| add feature autoencoder to NODE | NODE · Damped double pendulum | total MAE: 0.0269 -> 0.0602 | Hurt |
| use Hamiltonian bracket instead of NODE+AE | NODE+AE · MuJoCo HalfCheetah | relative error: 0.106 -> 0.0566 | Improved |
| use gradient bracket instead of NODE+AE | NODE+AE · MuJoCo HalfCheetah | relative error: 0.106 -> 0.105 | No clear effect |
| use metriplectic bracket instead of NODE+AE | NODE+AE · MuJoCo Swimmer | relative error: 0.0297 -> 0.0179 | Improved |
| use partially reversible double bracket instead of gradient bracket | gradient bracket · Computer (random split) | test accuracy: 78.1 -> 84.8 | Improved |
| use double bracket instead of gradient bracket | gradient bracket · Photo (random split) | test accuracy: 88.2 -> 92.4 | Improved |
| use Hamiltonian (fully reversible) bracket instead of gradient bracket | gradient bracket · Cora (Planetoid) | test accuracy: 79.9 -> 77.2 | Hurt |
| use metriplectic bracket instead of double bracket | double bracket · Cora (Planetoid) | test accuracy: 82.6 -> 57.4 | Hurt |
| double bracket vs GDE | GDE · Cora (Planetoid) | test accuracy: 83.8 -> 82.6 | No clear effect |
| double bracket vs GRAND-nl | GRAND-nl · CiteSeer (Planetoid) | test accuracy: 70.8 -> 74.2 | Improved |
Closest papers by meaning
- Graph Neural PDE Solvers with Conservation and Similarity-Equivariance (ICML 2024 · similarity 0.83)
- SEGNO: Generalizing Equivariant Graph Neural Networks with Physical Inductive Biases (ICLR 2024 · similarity 0.83)
- Physics-Informed DeepONets for drift-diffusion on metric graphs: simulation and parameter identification (ICML 2025 · similarity 0.83)
- Hyperbolic-PDE GNN: Spectral Graph Neural Networks in the Perspective of A System of Hyperbolic Partial Differential Equations (ICML 2025 · similarity 0.83)
- Supercharging Graph Transformers with Advective Diffusion (ICML 2025 · similarity 0.83)
- When Graph Neural Networks Meet Dynamic Mode Decomposition (ICLR 2025 · similarity 0.83)
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.
