Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Learning Differential Equations that are Easy to Solve
Jacob Kelly, Jesse Bettencourt, Matthew J Johnson, David K. Duvenaud
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 |
|---|---|---|---|
| Add second-total-derivative speed regularization | Neural ODE classifier · MNIST classification | NFE: 110 -> 104 | Improved |
| Add second-total-derivative speed regularization | Neural ODE classifier · MNIST classification | Loss: .0235 -> .0234 | Improved |
| Use Taylor-mode speed regularization instead of RNODE regularization | FFJORD · MINIBOONE | NFE: 182 -> 177 | Improved |
| Use Taylor-mode speed regularization instead of RNODE regularization | FFJORD · MINIBOONE | Loss: 9.77 -> 9.84 | Hurt |
| Add second-total-derivative speed regularization | FFJORD · MNIST density estimation | NFE: 149 -> 92 | Improved |
| Add second-total-derivative speed regularization | FFJORD · MNIST density estimation | Bits/Dim: 1.033 -> 1.039 | Hurt |
| Add trajectory-derivative speed regularization | Latent ODE · PhysioNet Challenge 2012 | Average NFE: 281 to 90 | Improved |
| Regularize the solver-matched second derivative rather than another order (ablation) | Neural ODE classifier · MNIST classification with a 2nd-order solver | Speed-performance tradeoff | Improved |
| Regularize the solver-matched third derivative rather than another order (ablation) | Neural ODE classifier · MNIST classification with a 3rd-order solver | Speed-performance tradeoff | Improved |
| Vary regularization order with a 5th-order solver (ablation) | Neural ODE classifier · MNIST classification with a 5th-order solver | Speed-performance tradeoff | Mixed |
| Increase regularization order for an adaptive-order solver (ablation) | Neural ODE classifier · MNIST classification with an adaptive-order solver | Speed-performance tradeoff | Improved |
| Use Taylor-mode AD rather than nested forward-mode differentiation | Higher-order derivative computation · Neural ODE training | Wall-clock computation time: an order of magnitude | Improved |
Closest papers by meaning
- Semi-supervised Learning of Dynamical Systems with Neural Ordinary Differential Equations: A Teacher-Student Model Approach (AAAI 2024 · similarity 0.87)
- How Deep Do We Need: Accelerating Training and Inference of Neural ODEs via Control Perspective (ICML 2024 · similarity 0.87)
- Learning to Solve Differential Equation Constrained Optimization Problems (ICLR 2025 · similarity 0.87)
- Neural Stochastic Flows: Solver-Free Modelling and Inference for SDE Solutions (NeurIPS 2025 · similarity 0.86)
- Simulation-Free Training of Neural ODEs on Paired Data (NeurIPS 2024 · similarity 0.86)
- An Efficient High-dimensional Gradient Estimator for Stochastic Differential Equations (NeurIPS 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.
