Research map / Probabilistic and scientific modeling / Dynamical systems
Neural Relational Inference for Interacting Systems
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 |
|---|---|---|---|
| Infer a latent interaction graph while learning the dynamics, rather than use correlation baselines | NRI (learned) · Springs, 5 objects | interaction recovery accuracy: 52.4 -> 99.9 | Helped |
| Infer a latent interaction graph while learning the dynamics, rather than use correlation baselines | NRI (learned) · Charged, 5 objects | interaction recovery accuracy: 55.8 -> 82.1 | Helped |
| Infer a latent interaction graph while learning the dynamics, rather than use correlation baselines | NRI (learned) · Kuramoto, 5 objects | interaction recovery accuracy: 62.8 -> 96.0 | Helped |
| Infer a latent interaction graph while learning the dynamics, rather than use correlation baselines | NRI (learned) · Springs, 10 objects | interaction recovery accuracy: 50.4 -> 98.4 | Helped |
| Infer a latent interaction graph while learning the dynamics, rather than use correlation baselines | NRI (learned) · Charged, 10 objects | interaction recovery accuracy: 51.4 -> 70.8 | Helped |
| Infer a latent interaction graph while learning the dynamics, rather than use correlation baselines | NRI (learned) · Kuramoto, 10 objects | interaction recovery accuracy: 59.3 -> 75.7 | Helped |
| Use the learned latent graph rather than a full graph | NRI (learned) · Springs, 5 objects; 20 prediction steps | MSE: 6.31e-3 -> 2.13e-5 | Helped |
| Use the learned latent graph rather than a full graph | NRI (learned) · Charged, 5 objects; 20 prediction steps | MSE: 9.24e-3 -> 7.06e-3 | Helped |
| Use the learned latent graph rather than a full graph | NRI (learned) · Kuramoto, 5 objects; 20 prediction steps | MSE: 8.96e-2 -> 3.26e-2 | Helped |
| Use the learned interaction graph rather than the true graph | NRI (learned) · Charged, 5 objects; 20 prediction steps | MSE: 5.71e-3 -> 7.06e-3 | Hurt |
| Dynamically re-evaluate the encoder at each test time step rather than keep the graph static | NRI on motion capture walking · CMU Motion Capture Database | predictive performance | Helped |
| Use NRI rather than the LSTM baseline for prediction | NRI on SportVU pick-and-roll · NBA SportVU | test MSE | Helped |
Closest papers by meaning
- A Graph Dynamics Prior for Relational Inference (AAAI 2024 · similarity 0.91)
- Online Relational Inference for Evolving Multi-agent Interacting Systems (NeurIPS 2024 · similarity 0.89)
- Interpretable Relational Inference with LLM-Guided Symbolic Dynamics Modeling (NeurIPS 2026 · similarity 0.88)
- Benchmarking Structural Inference Methods for Interacting Dynamical Systems with Synthetic Data (NeurIPS 2024 · similarity 0.87)
- Graph Mamba Operator: A Latent Simulator for Interacting Particle Systems (NeurIPS 2026 · similarity 0.86)
- PGODE: Towards High-quality System Dynamics Modeling (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.