Research map / Probabilistic and scientific modeling / Dynamical systems
Learning Continuous System Dynamics from Irregularly-Sampled Partial Observations
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 temporal self-attention over observations instead of taking the first observation (ablation) | LG-ODE · Springs interpolation 40% | MSE x 1e-2: 1.3017 -> 0.3350 | Helped |
| use temporal self-attention instead of mean pooling (ablation) | LG-ODE · Springs interpolation 40% | MSE x 1e-2: 0.3896 -> 0.3350 | Helped |
| use temporal self-attention instead of mean pooling (ablation) | LG-ODE · Motion extrapolation 40% | MSE x 1e-2: 1.3169 -> 1.2843 | Helped |
| add attention in temporal GNN (vs no att) (ablation) | LG-ODE · Springs interpolation 40% | MSE x 1e-2: 0.5145 -> 0.3350 | Helped |
| add learnable positional encoding (vs none) (ablation) | LG-ODE · Springs interpolation 40% | MSE x 1e-2: 0.4431 -> 0.3350 | Helped |
| learnable instead of fixed positional encoding (ablation) | LG-ODE · Springs interpolation 40% | MSE x 1e-2: 0.4285 -> 0.3350 | Helped |
| learnable instead of fixed positional encoding (ablation) | LG-ODE · Motion interpolation 40% | MSE x 1e-2: 0.4215 -> 0.4515 | Mixed |
| learnable instead of fixed positional encoding (ablation) | LG-ODE · Charged extrapolation 40% | MSE x 1e-2: 6.4520 -> 6.5320 | Mixed |
| learnable instead of fixed positional encoding (ablation) | LG-ODE · Springs extrapolation 40% | MSE x 1e-2: 1.7905 -> 1.7839 | Helped |
| add attention (vs no att) (ablation) | LG-ODE · Charged extrapolation 40% | MSE x 1e-2: 7.2958 -> 6.5320 | Helped |
| jointly model graph interaction in encoder (LG-ODE vs single-series Latent-ODE) | LG-ODE · Springs interpolation 40% | MSE x 1e-2: 0.5454 -> 0.3350 | Helped |
| end-to-end LG-ODE instead of two-step imputation then NRI | LG-ODE · Springs extrapolation 40% | MSE x 1e-2: 2.6638 -> 1.7839 | Helped |
Closest papers by meaning
- PGODE: Towards High-quality System Dynamics Modeling (ICML 2024 · similarity 0.89)
- Neural Persistence Dynamics (NeurIPS 2024 · similarity 0.88)
- Graph Mamba Operator: A Latent Simulator for Interacting Particle Systems (NeurIPS 2026 · similarity 0.88)
- CSG-ODE: ControlSynth Graph ODE For Modeling Complex Evolution of Dynamic Graphs (ICML 2025 · similarity 0.88)
- Learning system dynamics without forgetting (ICLR 2025 · similarity 0.87)
- Identifying latent state transitions in non-linear dynamical systems (ICLR 2025 · similarity 0.87)
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.