Companion code in JAX to the article preprint: Modelling pathwise uncertainty of Stochastic Differential Equations samplers via Probabilistic Numerics by Yvann Le Fay, Simo Särkkä and Adrien Corenflos.
This is a JAX implementation of 1.0 strongly convergent SDE schemes including novel Gaussian-based probabilistic SDE solvers.
- Classic SDE schemes: Euler-Maruyama, 1.5 Taylor-Itô
- Exotic Gaussian filtering SDE schemes including 1.0 strongly convergent scheme based on piecewise polynomial approximations of the Brownian motion. Can be used both for pathwise and moment computations.
- Euler ODE scheme.
- Extended Kalman filtering, with lower square root implementation.
See the scripts and tests folders for examples of usage.
Please refer to scripts/README.md for instructions on how to reproduce the results of the article.
This project is licensed under the MIT License.