Research map / Time series / Causal and generative time series
Learnable Path in Neural Controlled Differential Equations
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 |
|---|---|---|---|
| learn a task-specific latent path from the interpolated input path | NCDE · Character Trajectories, Speech Commands, PhysioNet Sepsis, and MuJoCo | accuracy, AUROC, or MSE: 0.992 ± 0.001 vs 0.988 ± 0.004; 0.922 ± 0.002 vs 0.912 ± 0.006; 0.908 ± 0.004 vs 0.886 ± 0.002; 0.022 ± 0.001 vs 0.028 ± | Helped |
| add the likelihood term to the learned-path training loss (ablation) | LEAP · Character Trajectories | accuracy: 0.973 -> 0.992 | Helped |
| add the likelihood term to the learned-path training loss (ablation) | LEAP · Speech Commands | accuracy: 0.921 -> 0.922 | Helped |
| add the likelihood term to the learned-path training loss (ablation) | LEAP · PhysioNet Sepsis | AUROC: 0.897 -> 0.908 | Helped |
| add the likelihood term to the learned-path training loss (ablation) | LEAP · MuJoCo | MSE: 0.024 -> 0.022 | Helped |
| use a small rather than medium LEAP model (ablation) | LEAP · PhysioNet Sepsis | AUROC: 0.908 -> 0.900 | Hurt |
| use a large rather than medium LEAP model (ablation) | LEAP · PhysioNet Sepsis | AUROC: 0.908 -> 0.894 | Hurt |
| add learnable layers to GRU-ODE instead of using the LEAP encoder-decoder (ablation) | GRU-ODE · Character Trajectories | accuracy: 0.875 vs 0.992 | Hurt |
| add learnable layers to ODE-RNN instead of using the LEAP encoder-decoder (ablation) | ODE-RNN · Speech Commands | accuracy: 0.428 vs 0.922 | Hurt |
| add learnable layers to NCDE instead of using the LEAP encoder-decoder (ablation) | NCDE · PhysioNet Sepsis | AUROC: 0.881 vs 0.908 | Hurt |
| add learnable layers to NCDE instead of using the LEAP encoder-decoder (ablation) | NCDE · MuJoCo | MSE: 0.037 vs 0.022 | Hurt |
| vary the forecasting output sequence length from 1 to 20 (ablation) | LEAP · MuJoCo | MSE | Helped |
Closest papers by meaning
- Controlled Differential Equations on Long Sequences via Non-standard Wavelets (ICML 2023 · similarity 0.91)
- Log Neural Controlled Differential Equations: The Lie Brackets Make A Difference (ICML 2024 · similarity 0.90)
- On the Generalization and Approximation Capacities of Neural Controlled Differential Equations (ICLR 2024 · similarity 0.90)
- Faithful Embeddings of Irregular and Asynchronous Data for Online Log-NCDEs (NeurIPS 2026 · similarity 0.86)
- Symbolic Neural Ordinary Differential Equations (AAAI 2025 · similarity 0.86)
- ControlSynth Neural ODEs: Modeling Dynamical Systems with Guaranteed Convergence (NeurIPS 2024 · similarity 0.85)
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.