Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Locally Regularized Neural Differential Equations: Some Black Boxes were meant to remain closed!
Avik Pal, Alan Edelman, Christopher Vincent Rackauckas
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 |
|---|---|---|---|
| Use uniform random time-point sampling for local error regularization | Neural ODE · MNIST | Prediction Time (s / batch): 0.088 ± 0.020 -> 0.046 ± 0.002 | Improved |
| Sample regularization points from solver time points | Neural ODE · MNIST | Testing NFE: 303.559 ± 3.194 -> 182.849 ± 1.578 | Improved |
| Use local regularization | Neural ODE · Physionet 2012 | Prediction Time (s / batch): 0.16 ± 0.01 -> 0.09 ± 0.00 | Improved |
| Use local regularization with uniform time-point sampling | Latent NODE · Physionet 2012 | Test Loss (× 10^-3): 3.41 ± 0.10 -> 3.64 ± 0.07 | Hurt |
| Use local regularization with uniform time-point sampling | NSDE · MNIST | Prediction Time (s / batch): 0.51 ± 0.07 -> 0.39 ± 0.10 | Improved |
| Use local regularization, sampling from solver time points | Neural ODE · CIFAR10 Standard Neural ODE | Testing NFE: 115.315 ± 12.136 -> 81.301 ± 1.255 | Improved |
| Use local error regularization | Multiscale Neural ODE · CIFAR10 Multi-Scale | Prediction Time (s / batch): 0.170 ± 0.005 -> 0.175 ± 0.019; 0.170 ± 0.005 -> 0.163 ± 0.015 | Mixed |
| Regularize solver error estimates using local sampling rather than global accumulation | Neural Differential Equations · Across ODE and SDE experiments | Testing NFE: The abstract reports ×change{0.556}{0.733}. | Improved |
| Use stiffness estimates for local regularization (ablation) | Neural Differential Equations · Experiments | performance | Hurt |
Closest papers by meaning
- How Deep Do We Need: Accelerating Training and Inference of Neural ODEs via Control Perspective (ICML 2024 · similarity 0.89)
- Breakeven complexity: A new perspective on neural partial differential equation solvers (NeurIPS 2026 · similarity 0.89)
- Global Convergence in Neural ODEs: Impact of Activation Functions (ICLR 2025 · similarity 0.88)
- On the Generalization and Approximation Capacities of Neural Controlled Differential Equations (ICLR 2024 · similarity 0.88)
- Improving Neural ODE Training with Temporal Adaptive Batch Normalization (NeurIPS 2024 · similarity 0.88)
- Stable Neural Stochastic Differential Equations in Analyzing Irregular Time Series Data (ICLR 2024 · similarity 0.88)
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.
