Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Stable Implementation of Probabilistic ODE Solvers
Nicholas Krämer, Philipp Hennig
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 stable implementation (preconditioning, square-root, Taylor-mode init) with EK0 (ablation) | probabilistic ODE solver EK0 · Logistic ODE, tol 1e-5 | solver success by order nu: fails from nu=5 -> succeeds up to nu=11 | Improved |
| use stable implementation with EK1 (ablation) | probabilistic ODE solver EK1 · Logistic ODE, tol 1e-5 | solver success by order nu: fails from nu=6 -> succeeds up to nu=11 | Improved |
| add proposed coordinate change instead of no preconditioning (ablation) | Q_h · Process noise covariance, nu=5 | log10 cond(Q): 43.7 -> 7.2 | Improved |
| add proposed coordinate change instead of no preconditioning (ablation) | Q_h · Process noise covariance, nu=11 | log10 cond(Q): 79.9 -> 16.2 | Improved |
| proposed coordinate change instead of Nordsieck coordinates (ablation) | Q_h · Process noise covariance, nu=11 | log10 cond(Q): 17.4 -> 16.2 | Improved |
| proposed coordinate change instead of Nordsieck coordinates (ablation) | Q_h · Process noise covariance, nu=11 | log10 rho: 5.6 -> 1.4 | Improved |
| proposed coordinate change instead of Nordsieck coordinates (ablation) | Q_h · Process noise covariance, nu=1 | log10 cond(Q): 1.3 -> 1.3 | No clear effect |
| proposed coordinate change instead of no preconditioning (ablation) | Q_h · Process noise covariance, nu=3 | log10 min eigenvalue: E -> -4.0 | Improved |
| increase order nu up to 11 with EK1 | stable EK1 · Lotka-Volterra, three-body | convergence rate: at least h^nu, observed h^(nu+1) | Improved |
| EK0 instead of EK1 on stiff problems | probabilistic ODE solver · stiff van der Pol | solvability | Hurt |
| increase EK0 order beyond 5 | EK0 · three-body | adaptive step-size efficiency | Hurt |
Closest papers by meaning
- Stable and Scalable Probabilistic Numerical Solvers for Stiff and High-Dimensional ODEs (NeurIPS 2026 · similarity 0.94)
- Probabilistic ODE Solutions in Millions of Dimensions (ICML 2022 · similarity 0.92)
- Parallel-in-Time Probabilistic Numerical ODE Solvers (JMLR 2024 · similarity 0.90)
- Diffusion Tempering Improves Parameter Estimation with Probabilistic Integrators for Ordinary Differential Equations (ICML 2024 · similarity 0.85)
- Affine Tracing: A New Paradigm for Probabilistic Linear Solvers (NeurIPS 2026 · similarity 0.84)
- Stability-Informed Initialization of Neural Ordinary Differential Equations (ICML 2024 · similarity 0.84)
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.
